# Home Welcome to your team’s developer platform

Welcome to the IFS Loops Knowledge Base. Select your IFS Loops product to learn more.

Cover image
Agent StudioA no-code platform for building, deploying, and managing AI agents that automate complex business workflows./spaces/RIbp8Kp04VLXld01gmAs/pages/njH9eu74kH4nq2PaW1uB/files/WrFzE9NC7uthj0TWfKnV
Digital WorkersA suite of AI-powered digital workers that automate high-volume operational processes across supply chain, procurement, and inventory management./spaces/wtaSDVCrMbb4Gnba4hQA/pages/9PmfeU7SRvkGyD3TWF6t/files/lXvnOiAzvD0GGQn8Uk15
CX PlatformAI-powered tools that automate and enhance customer interactions, agent support, and contact center performance management./spaces/o3WqE1JEUAvmvC1DxJHK/files/4u5zGVHgZoh8F8fXsEku

# IFS Loops Digital Workers Welcome to your team’s developer platform

IFS Loops Digital Worker Knowledge Base

Welcome! Explore our guides and resources to get the most out of IFS Loops Digital Workers, purpose-built AI that autonomously executes complex processes across your enterprise systems so your team can focus on the work that matters most.

Cover image
Digital Worker guidesGuides and resources for every Digital Worker in your toolkit./pages/j7NdAgY1nNSl6TBYC53C/files/YX0vw9x8NaieNDLZilYR
How-to guidesPractical, step-by-step guidance across a range of tasks./pages/2Qs8NANzzJTvoXvABd7B/files/uC1RptEfw6kewkOX3sMG
IntegrationsConnect seamlessly with 65+ enterprise tools across CRM, ticketing, and more./spaces/RIbp8Kp04VLXld01gmAs/pages/DBHws38eO8AG9o0lUs7k/files/juHGw45H4mltsa8beSnG
Release notesStay up to date with the latest product updates and feature enhancements./spaces/zpOPAqpE6CRlx7tqonUl/pages/MIGOOkHLmdctGK1QVdaK/files/eenjQFp0yQn9hRm2swHK
# Digital Worker guides Each Digital Worker is purpose-built for a specific role in your operations. Select one below to explore what it does and how it works.
Cover image
Customer Order ManagerEnd-to-end customer order automation

:inboxes:

/pages/IsQJTcYhcRABFKN6z8QN/files/15w6xTOtT6axsqrEbeVB
Dispatcher AssistantManage real-time service disruptions/pages/FCmIVybJTut60GYXmpqs/files/EKDYfe4y4Fv5R1qK0bAM
Inventory ReplenisherReal-time demand forecasting & restocking

:pallet-boxes:

/pages/okYnI9Us0gmfyOX8qod0/files/vxYmcZhaUG2oekE39dL1
Knowledge ManagerDeliver real-time, contextual knowledge to service teams across any channel/pages/2PDk26vNOUZ1UgJAgLKD/files/wyiDW8wZjmdUuhZEvzR7
Material ReplenisherAutomated parts sourcing & allocation

:wand-sparkles:

/pages/WYmUGFCEVZNFVqanWZFz/files/jpANWwnvxU7EoiWtd1LA
Supplier Invoicing ManagerAutomated invoice intake, validation, and exception handling/pages/JBYr5n81oYGBzkqEYzJd/files/TaTXHypcnQ6Q3ElqbOvk
Supplier Order ManagerEfficiently manage supplier order confirmations and communication/pages/HcXjh7ViaII5j0lp3jw1/files/UeFYDTRtV6rq7OA3VRqL
# Customer Order Manager Handling customer orders takes focus, but repetitive intake work shouldn’t slow teams down. This knowledge base guides you through using the Customer Order Manager, from initial setup to handling customer orders efficiently in daily operations. ### Jump right in
Cover image
Watch & learnSee this Digital Worker in action./files/IniyzVQmRw3Net0b8IkX/pages/jaQwgyEmzj23RB2KRoPf
Core conceptsEssential context for understanding how it all works./files/fZNxfxVnHxx6iNh5M9k9/pages/BOzsbxQrwDa0q0Euqxwb
FAQQuick answers to common questions./files/RBUXYX1hxsMuVshLZ5Ar/pages/NbRBRd9HUGzhU6lNMxdX
How-to guidesPractical, step-by-step guidance across a range of tasks/files/uC1RptEfw6kewkOX3sMG/pages/2Qs8NANzzJTvoXvABd7B
# Watch & learn Watch the video below to see this Digital Worker in action {% embed url="" %} # Core concepts Explore core concepts to build a solid understanding of the Digital Worker. These articles will give you the context needed to get the most out of your Digital Worker.
Cover image
FundamentalsLearn the full scope of what's possible/files/PKWzoCypd9fEWl1gt5o8/pages/MhbGFeYmDjLMNYUD0UJe
Architecture & capabilitiesDiscover the components and integrations/files/MgisEby9ZvW43Wybsymq/pages/5dz10kLmW9tEmCkdIaMq
Real-world scenariosPractical examples of the platform in action across common business use cases./files/I8b6oW47o0Zv7grfRIxS/pages/p4Oq253sLWkGsqYU8cLv
# Fundamentals ## Overview The Customer Order Manager is an AI-powered Digital Worker designed to free order processing teams from manual data entry so they can focus on the work that actually requires human expertise. It handles incoming customer orders, extracts critical information from various document formats, validates order details against IFS Cloud, and creates complete order records automatically. By transforming unstructured, manual order intake into a fast, reliable, and error-resistant process, the Customer Order Manager gives teams their time back for building customer relationships, resolving complex issues, and managing exceptions that require genuine judgment. ## Key benefits **Universal format handling:** Processes orders arriving in any format, PDFs, images, Excel sheets, emails, EDI, eliminating format fragmentation and standardizing intake across all channels. **Accuracy and speed combined:** Intelligent validation eliminates part number mismatches, unit conversion errors, and duplicate orders, while delivering instant customer acknowledgment. **Error prevention built-in:** The system cross-checks part numbers, quantities, prices, and delivery details against system records before creating any order. Issues are flagged before they become problems. **Seamless scalability:** Handles high-volume and peak-season periods without bottlenecks, processing orders consistently regardless of volume. **Human expertise elevated:** Routine order creation is handled automatically. Teams focus on exceptions, customer relationships, and decisions that require business judgment. ## The Challenge It Solves Customer order management is often manual and fragmented. Teams must interpret incoming requests, validate details, and create orders across multiple systems. This leads to delays in order processing, errors in data entry, inconsistent customer communication, and limited visibility into order status. In practice, this means teams are spending time on tasks like opening email attachments, re-entering data into ERP systems, manually validating part numbers, converting units, chasing clarifications, and sending acknowledgment emails. High order volumes and inconsistent incoming formats compound the problem, creating bottlenecks during peak periods and increasing the risk of errors that require costly correction. ## Before and after ### Before Before the Customer Order Manager, order processing requires manual effort at every stage: {% stepper %} {% step %} Process inbox and review purchase orders via email, EDI, or portal {% endstep %} {% step %} Extract order details manually {% endstep %} {% step %} Validate customer pricing and terms {% endstep %} {% step %} Check availability constraints {% endstep %} {% step %} Handle discrepancies via email back-and-forth {% endstep %} {% step %} Send acknowledgments {% endstep %} {% step %} Update ERP {% endstep %} {% step %} Follow up on exceptions {% endstep %} {% endstepper %} This manual process introduces data entry errors, inconsistent validation, slow processing cycles, and a poor customer experience during the wait for confirmation. ### After With the Customer Order Manager, the same cycle is compressed into four automated steps: {% stepper %} {% step %} Extract unstructured purchase orders from multiple channels {% endstep %} {% step %} Validate pricing and terms, handle discrepancies {% endstep %} {% step %} Check availability and handle exceptions {% endstep %} {% step %} Update ERP and send automated acknowledgments {% endstep %} {% endstepper %} Routine orders are processed without any human involvement. Exceptions are routed to the right person with full context already attached. ## How it works The Customer Order Manager executes a comprehensive automated workflow that handles the complete order lifecycle from intake to confirmation. **Extraction and initial setup:** The Digital Worker applies extraction rules to pull relevant information from incoming order documents. It retrieves default values such as site ID and coordinator information, identifies the customer by searching for and retrieving the customer ID from the system, and applies site mapping by checking if the customer is registered to a specific site ID using predefined dictionary mapping. **Validation and cross-referencing:** Before creating the order, the system verifies the customer PO number doesn't already exist in IFS Cloud to prevent duplicates, retrieves customer-specific order defaults, and resolves any differences between the customer's part numbers and internal sales part numbers. **Part number resolution:** The system cross-references customer part numbers with sales part numbers and retrieves conversion factors to translate between different quantity representations. This handles scenarios where customers count parts in bulk, such as sets of 12, while the system tracks individual units. **Address matching:** The Digital Worker fetches all addresses stored for the customer and matches the delivery address from the purchase order with system records to ensure accurate fulfilllment. **Order creation:** The system creates the complete customer order in IFS Cloud with all validated information and updates the part number registry if new parts have been added. **Customer communication:** The Digital Worker sends acknowledgments and confirmations back to customers, ensuring timely communication throughout the process. # Architecture & capabilities ## Overview The Customer Order Manager is built on a set of integrated agentic skills coordinated by a Supervisor Agent. This article covers the core skills that make up the Digital Worker, how its agent modes and configuration options work, the full order management workflow, integration points, and how it determines when to proceed automatically versus when to involve a human. ## Agentic skills The Customer Order Manager operates through five core skills. Each handles a distinct part of the order management process. #### Customer communication Handles all order-related conversations with customers, including acknowledgments, status updates, and clarifications. This skill sends confirmations when orders are received and processed, and drafts clarification requests when information is missing or unclear. All customer-facing communication is managed through this skill, ensuring a consistent and timely response regardless of order volume. #### Order processing Interprets incoming orders and converts them into validated purchase order confirmations with correct details and validations. This skill parses order content from emails, PDFs, and other document formats, structures the data, validates completeness, and prepares it for ERP entry. It handles unstructured formats and applies extraction rules to reliably pull part numbers, quantities, prices, and delivery information. #### Asset discovery Identifies the relevant assets, configurations, or product context needed to fulfilll the customer order accurately. This skill validates parts, sizes, and units against system records, ensuring that what the customer has ordered maps correctly to what exists in IFS Cloud before any order is created. #### Quote evaluation Reviews pricing and quote details to ensure the order aligns with agreed commercial terms before processing. This skill checks customer-specific pricing, discounts, and contract terms, flagging any discrepancies between the incoming order and the terms on record. #### Notifications Keeps teams informed by sending timely alerts and updates when actions, approvals, or exceptions are required. When an order is processed successfully, teams receive confirmation via Microsoft Teams. When an exception is detected, the notification includes full context so the team can make a decision quickly. ## Agent modes and configuration The Customer Order Manager offers two agent configurations to match the complexity of different order types. **Basic agent mode** uses standard language models for straightforward order processing with clear part numbers and standard formats. This mode is suitable for high-volume, well-structured orders. **Reasoning mode** employs advanced reasoning models for complex scenarios requiring deeper analysis, unclear formats, or sophisticated validation logic. This mode is appropriate when orders involve ambiguous formatting, non-standard part references, or multi-step validation requirements. In Reasoning Mode, administrators can configure: * **Reasoning effort:** Controls how much computational effort the agent applies to problem-solving and validation * **Verbosity:** Adjusts how detailed the agent's responses and logging will be, supporting transparency and debugging * **Concurrent processing:** Enables parallelization of tool execution to improve performance when handling multiple orders simultaneously ## Order management workflow The following table outlines the full order management workflow from initial intake through to customer communication. | Order Intake & Parsing | Parses orders from email, EDI, and portals; extracts customer, items, quantity, price, and dates; handles unstructured formats | Structures order data, validates completeness, triggers order workflow | Email / EDI / IFS | | --------------------------------- | ------------------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------- | ----------------- | | Order Validation & Contract Check | Validates pricing, checks customer terms and discounts, identifies inconsistencies | Flags discrepancies, applies validation rules, routes for exception or auto-processing | IFS / Teams | | Availability & Feasibility Check | Checks stock availability, identifies constraints such as lead time or capacity, detects fulfilllment risks | Confirms feasibility, flags constraints, triggers alternative flows if needed | IFS | | Order Creation & Processing | Validates structured order data, ensures ERP readiness | Creates order in IFS Cloud, ensures data consistency, triggers downstream processes | IFS | | Exception Handling & Resolution | Identifies mismatches in price, date, or quantity; detects incomplete orders | Triggers human-in-the-loop review, suggests resolution options, resumes flow after decision | Teams / Email | | Customer Communication | Detects need for acknowledgment or clarification, identifies delays or issues | Sends order confirmations, drafts clarification emails, provides status updates | Email | ## Tools and integration layer The Customer Order Manager uses a comprehensive set of tools to interact with IFS Cloud and connected systems: * **Get Defaults:** retrieves site ID and coordinator defaults * **Get Customer ID:** searches and matches customer names to retrieve customer IDs * **Get Customer Bases:** fetches customer-specific defaults needed for order creation * **Get Addresses:** retrieves stored customer addresses for validation and matching * **Check PO Number:** validates whether a customer PO number has been used previously * **Get Sales Part Numbers:** retrieves cross-reference information for part number resolution * **Get Part Information:** fetches detailed information for each part including conversion factors * **Create and Update Order:** generates the customer order record in IFS Cloud * **Update Entities:** adds new parts to the system when necessary ## Integration points **Email Systems:** Monitors Microsoft Outlook and Gmail inboxes continuously, processing incoming customer orders with PDF, image, and Excel attachments in real time. **ERP Platforms:** Integrates directly with IFS Cloud to read customer records, validate part numbers and pricing, and write complete order details automatically. **Collaboration Platforms:** Uses Microsoft Teams, Slack, OneNote, and Google Drive for exception notifications, approvals, and documentation. **CRM Systems:** Connects with Salesforce, HubSpot, and Qoncierge to maintain customer relationship context and communication history. **Document Processing:** Handles PDFs, images (JPG, PNG), Excel sheets, and EDI with high extraction accuracy using OCR and AI-powered parsing. ## Human-in-the-loop design The Customer Order Manager is designed so that routine processing happens automatically and human judgment is applied where it's genuinely needed. The system proceeds independently when: * Part numbers are valid and exist in the system * Quantities are standard and delivery requirements are clear * Customer information matches existing records * The order is complete with no exceptions or discrepancies Human input is required when: * Part numbers are invalid or don't exist in the system * Information is unclear or missing and requires clarification * Only some line items in an order are valid * Orders involve pricing discrepancies or contract term conflicts * Business judgment is required for high-value or complex decisions When an exception is detected, the system identifies the specific issue, sends a detailed notification via Microsoft Teams with complete context, allows team members to provide guidance directly in the Teams channel, and executes the approved action immediately once a decision is made. ## Testing and validation The platform includes an evaluation environment for testing and development. Teams can input sample email bodies with realistic content, attach PDF documents containing customer order information, trigger the agent and observe processing in real time, and monitor all tool calls and interactions. Debug mode provides detailed visibility into the inputs and outputs of each tool call. Advanced view breaks down each operation into form input, API task, and response processing components for complete transparency into how each step is being handled. # Real-world scenarios ## Overview This article walks through two practical scenarios that illustrate how the Customer Order Manager handles incoming purchase orders. The first covers a clean, straightforward order that processes without any human intervention. The second covers an order with invalid line items, showing how the Digital Worker detects the issue, escalates to a team member, and executes the corrected order once a decision is made. ### Scenario 1: Standard order processing This scenario covers the most common type of interaction, a customer sends a purchase order with valid part numbers, standard quantities, and clear delivery requirements. Everything matches what's in the system. **What triggers it:** A customer sends an email to the order manager's monitored inbox with a purchase order attached as a PDF. **What the Digital Worker does:** 1. The Customer Communication skill detects the incoming email, identifies it as a purchase order, and acknowledges receipt to the customer. 2. The Order Processing skill extracts all fields from the PDF, customer details, part numbers, quantities, prices, and delivery dates, and structures them for validation. 3. The Asset Discovery skill validates the parts, sizes, and units against IFS Cloud to confirm everything is accurate and matches system records. 4. The order is created in IFS Cloud with complete details and released into the processing workflow. 5. The Notifications skill sends confirmation to the team via Microsoft Teams and a confirmation back to the customer that their order has been received and is being fulfillled. **Result:** The complete order is created in IFS Cloud with no manual intervention. The team receives immediate visibility via Teams. The customer receives a confirmation without having to follow up. The entire process, from email receipt to ERP entry to customer acknowledgment, happens automatically. ### Scenario 2: Order with invalid line items This scenario covers an order where not all line items are valid. The customer submits an order with three line items, but only one part number exists in the system. The other two are invalid. #### Step 1: Order received and processed The customer sends their purchase order via email, unaware that two of the three part numbers cannot be found in IFS Cloud. The Digital Worker picks up the email, reads through it, and begins processing the attached PDF, cross-referencing every line item against the part catalog in IFS Cloud at the same time. #### Step 2: Exception detected The system identifies that two of the three part numbers are invalid. Rather than making assumptions, processing bad data into the system, or rejecting the entire order, the Digital Worker flags the issue and sends a notification via Microsoft Teams. The notification includes full context: which line items are valid, which are not, and what the issue is. The team has everything they need to make a decision without doing any additional investigation. #### Step 3: Human decision made A team member reviews the exception in Teams and instructs the Digital Worker to proceed, creating the order for the one valid line item and holding the others pending clarification with the customer. This decision takes seconds, not the 10 to 15 minutes it would have taken to manually process the entire order from scratch and then discover the errors. #### Step 4: Order created The Digital Worker executes the instruction immediately. The order is created in IFS Cloud with only the valid line item, with correct quantity and all other details exactly as specified. The team receives confirmation that the order has been created. **Result:** The Customer Order Manager didn't blindly process incorrect data or reject the order entirely. It identified exactly what was wrong, surfaced the issue to the right person with full context, and executed precisely what was decided, leaving the team in control of the exception while handling all the processing work automatically. ## What these scenarios illustrate **Clean orders require no human involvement.** When an order is complete and valid, the Digital Worker handles the entire process end to end, extraction, validation, ERP entry, and customer confirmation, without the team needing to look at it. **Exceptions are caught before they cause problems.** The system detects issues at the validation stage, before any bad data enters IFS Cloud. Teams are notified with enough context to make a fast, informed decision. **Partial orders can be actioned.** The Digital Worker doesn't treat an order as all-or-nothing. Valid line items can be processed independently while exceptions are handled separately, keeping fulfilllment moving without waiting for every issue to be resolved. **Teams stay in control of decisions, not data entry.** The human role in the process is judgment, deciding what to do with exceptions, how to communicate with customers, and what the right business call is. The Digital Worker handles all of the processing that surrounds those decisions. # FAQ ### Capabilities & functionality **Q: What does Customer Order Manager do for me?** A: It streamlines customer communications, automatically updates sales orders, manages delivery requests, and sends confirmations — reducing manual effort so you can focus on customer satisfaction. **Q: How quickly does Customer Order Manager respond to customer requests?** A: In near real-time. As soon as a customer email or portal message arrives, Customer Order Manager scans it, extracts key details, and updates your ERP or CRM system, often before your team sees the message. **Q: How does Customer Order Manager know what to update?** A: It parses customer communications (emails, attachments, or forms) to capture order details and cross-checks them against your existing sales orders for accuracy. **Q: What if a customer changes multiple details at once?** A: Customer Order Manager can process combined updates — for example, a shipping address change and an updated quantity — applying them accurately and keeping a full record of the original request. **Q: Does Customer Order Manager support global customers?** A: Yes. It supports multilingual parsing, international shipping details, multiple currencies, and region-specific tax rules to ensure orders are processed correctly. ### Control & collaboration **Q: Will customers realize they're interacting with a digital worker?** A: No. Customer Order Manager uses professional business communication templates, so responses appear to come from your customer service or sales team. **Q: Can I override Customer Order Manager's decisions?** A: Yes. You can define thresholds for automation and set rules requiring human review — for example, high-value order changes, unusual requests, or special discounts. **Q: What happens if a customer doesn't confirm their order?** A: Customer Order Manager automatically follows up with reminder messages. If the customer remains unresponsive, it escalates the case to a team member. **Q: How will this affect my role?** A: Instead of manually updating orders and chasing confirmations, you'll spend more time resolving complex issues, improving customer experience, and growing customer relationships. ### Integration & technology **Q: How does Customer Order Manager connect to ERP or CRM systems?** A: It uses standard APIs to validate and update orders, ensuring customer requests flow directly into your core systems. **Q: Which APIs does Customer Order Manager rely on?** A: Core APIs handle sales order creation, updates, confirmations, and document retrieval to maintain accurate and reliable records. **Q: Can Customer Order Manager integrate with tools beyond ERP?** A: Yes. It can connect with Outlook, Gmail, Teams, and other communication platforms to centralize customer interactions across channels. ### Security & reliability **Q: How are customer interactions tracked?** A: Every message sent, order updated, and confirmation logged is written back to your ERP or CRM history. Entries are stamped with the digital worker's ID for full transparency. **Q: How is customer data protected?** A: Sensitive customer and order data stays within your ERP, CRM, and email environments. Customer Order Manager uses secure APIs and logs all transactions for compliance. **Q: What happens if ERP or CRM systems are unavailable?** A: Customer Order Manager queues updates and retries once systems are restored. Urgent customer communications are still drafted and logged, then reconciled when connectivity returns. ### Customization & performance **Q: Can Customer Order Manager adapt to our unique customer workflows?** A: Yes. It supports configuration of custom rules, triggers, and approval steps tailored to your ERP or CRM setup and customer engagement practices. **Q: How is performance measured?** A: Typical KPIs include order response time, percentage of automated customer requests processed, reduction in manual errors, and time saved per week. These are tracked via dashboards. # Case studies See how our customers are using this Digital Worker to transform their operations. # Dispatcher Assistant Managing dispatch operations takes coordination, but repetitive scheduling work shouldn't hold teams back. This knowledge base guides you through using the Dispatcher Assistant to streamline dispatch workflows in daily operations. ### Jump right in
Cover image
Watch & learnSee this Digital Worker in action./files/IniyzVQmRw3Net0b8IkX/pages/6ZUqaqGkqgWEUxRiaho7
Core conceptsEssential context for understanding how it all works./files/fZNxfxVnHxx6iNh5M9k9/pages/JuZKHvFzDxWA8pJHJeHO
FAQQuick answers to common questions./files/RBUXYX1hxsMuVshLZ5Ar/pages/ygCShnK7gyFEGPe5pz9u
How-to guidesPractical, step-by-step guidance across a range of tasks/files/uC1RptEfw6kewkOX3sMG/pages/2Qs8NANzzJTvoXvABd7B
# Watch & learn Watch the video below to see this Digital Worker in action {% embed url="" %} # Core concepts Explore core concepts to build a solid understanding of the Digital Worker. These articles will give you the context needed to get the most out of your Digital Worker.
Cover image
FundamentalsLearn the full scope of what's possible/files/PKWzoCypd9fEWl1gt5o8/pages/6vcM8XoRM71ifJOLyEL5
Architecture & capabilitiesDiscover the components and integrations/files/MgisEby9ZvW43Wybsymq/pages/iyqAQIAoFONkNdt18GmC
Real-world scenariosPractical examples of the platform in action across common business use cases./files/I8b6oW47o0Zv7grfRIxS/pages/CXlUHt44QpZvnejx4YW6
# Fundamentals ## Overview The Dispatcher Assistant is an AI-powered Digital Worker designed to keep field service operations stable when conditions aren't. It monitors live schedules and execution signals continuously, identifies exceptions and SLA risks early, while corrective options still exist, and generates ranked reallocation options with clear reasoning behind each one. Low-risk actions can proceed automatically. Higher-impact decisions are escalated to dispatchers with everything needed to confirm quickly. Customer communication is handled throughout, keeping all parties informed without adding to the dispatcher's workload. By converting reactive, manual exception handling into a structured, proactive process, the Dispatcher Assistant gives dispatch teams time back for the complex trade-offs, difficult customer conversations, and judgment calls that genuinely require human experience. ## Key Benefits **Early exception detection:** SLA risks and schedule disruptions are identified as they develop, not after a breach has already occurred, giving dispatchers time to act while options are still available. **Ranked reallocation options:** When an exception is detected, the system evaluates engineer availability, skill match, travel feasibility, and downstream impacts, then surfaces ranked options with the reasoning behind each one, so dispatchers make informed decisions rather than quick guesses. **Automated customer communication:** Customers are notified automatically when schedules change, appointments are renegotiated via email or SMS, and confirmations are handled end to end, without dispatcher involvement for routine cases. **Consistent exception decisions:** All disruptions follow the same structured, explainable flow aligned with PSO logic, eliminating the variability that comes from relying on individual dispatcher experience across regions and teams. **Scalable disruption management:** The same approach handles both individual exceptions and large-scale regional events such as weather or outages, without requiring additional dispatcher headcount. **Full traceability:** Every exception, recommendation, and schedule change is tracked and auditable end to end. ## The challenge it solves Dispatch teams operate in a constant state of disruption. Delays, cancellations, overruns, and last-minute customer changes force dispatchers into reactive firefighting. SLA risks are often identified too late, schedules are manually reshuffled mid-day, and communication across customers, engineers, and systems becomes fragmented. At scale, regional events amplify the problem, creating operational bottlenecks and inconsistent outcomes. In practice, dispatchers are spending 40 to 60% of their time on exception management: manually reassigning jobs, calling or emailing customers to reschedule, spending 10 to 20 minutes per disrupted job on coordination, managing bulk disruption events without a structured process, and relying heavily on individual experience to make reallocation decisions. Schedule changes and customer communication are handled separately, creating gaps and inconsistency. ## Before and After ### Before Before the Dispatcher Assistant, handling a schedule exception requires manual effort at every stage: {% stepper %} {% step %} Monitor engineer delays, overruns, and incoming exceptions {% endstep %} {% step %} Review job details, SLA commitments, and the current schedule {% endstep %} {% step %} Assess impact across downstream appointments {% endstep %} {% step %} Identify conflicts and prioritization issues {% endstep %} {% step %} Manually explore reallocation or rescheduling options {% endstep %} {% step %} Contact customers to renegotiate appointments {% endstep %} {% step %} Update the schedule and systems with agreed changes {% endstep %} {% endstepper %} This process is slow, inconsistent across teams, and limited in its ability to account for downstream impacts before changes are made. ### After With the Dispatcher Assistant, the same process runs in five automated steps:
{% stepper %} {% step %} Detect exceptions in real time, monitors delays, overruns, and regional events automatically {% endstep %} {% step %} Evaluate impact on schedule and SLA, identifies affected appointments using PSO and business rules {% endstep %} {% step %} Generate optimized resolution options, proposes engineer swaps, resequencing, or alternative slots {% endstep %} {% step %} Orchestrate customer negotiation, engages customers via email or SMS, presents options, manages responses {% endstep %} {% step %} Execute and update systems, confirms the new appointment, updates IFS Cloud, logs the outcome and audit trail {% endstep %} {% endstepper %} ## How It Works The Dispatcher Assistant continuously monitors live schedules and execution signals in PSO, interpreting customer and system-driven exceptions and assessing their impact on time-to-breach, SLA exposure, and engineer capacity. **When an exception is detected:** the system evaluates the situation using real-time availability, engineer capability and location, SLA and priority constraints, travel optimization, and customer history, accounting for downstream impacts before any recommendation is made. **Best-fit appointment options are generated:** ranked by SLA exposure, travel efficiency, and priority. For routine, low-risk exceptions the system can proceed automatically. For higher-impact or lower-confidence decisions, the exception is escalated to the dispatcher with ranked options and clear reasoning, everything needed to confirm quickly without additional investigation. **Customer communication is managed throughout:** proposing new appointment windows, handling back-and-forth responses, refining slots based on customer feedback, and confirming the agreed appointment across all stakeholders. **All decisions are revalidated and logged before commit:** maintaining full oversight and auditability. ## What Triggers the Dispatcher Assistant The Dispatcher Assistant responds to a wide range of operational exceptions that break or threaten the live schedule: **Engineer availability shifts:** engineer calls in sick, is stuck in traffic, experiences a vehicle breakdown, exceeds driving hours, or has a certification lapse. **Job duration drift:** a job overruns its expected duration due to asset complexity, unexpected parts requirements, or safety escalations. **SLA and priority changes:** a customer escalates, a VIP account is prioritized, an asset criticality is reclassified, or a regulatory deadline approaches. **Parts and material disruptions:** a required part is unavailable at van stock, a transfer is delayed, or an incorrect part was dispatched. **Customer behavioral changes:** a customer cancels last minute, restricts site access, or requests a different appointment window. **Environmental and external event:** severe weather, flooding, transport strikes, or regional power outages requiring bulk rescheduling. **System and data anomalies:** duplicate work orders, missing SLA data, or scheduling rule misconfigurations.
# Architecture & capabilities ## Overview The Dispatcher Assistant is built on four agentic skills coordinated by a Support Agent. It operates as a real-time exception orchestrator, running continuously alongside PSO, monitoring live operational signals, and managing the full exception response cycle from detection through to customer confirmation and system update. This article covers the core skills, the full dispatch management workflow, integration points, and how the system determines when to act autonomously versus when to escalate to a dispatcher. ## Agentic Skills #### Customer Exception Parsing & Intent Detection Understands and interprets customer communications and operational signals to accurately identify service disruptions and required actions. This skill distinguishes between different exception types, cancellation, reschedule request, site access issue, job scope change, and extracts relevant context such as asset, location, work order, and urgency level. It classifies the exception and triggers the appropriate workflow, including highlighting SLA exposure risk before any action is taken. #### Engineer & Resource Reallocation Analysis Analyzes engineer availability and workload to recommend optimal reallocations and minimize service disruption. This skill evaluates engineer skill match against job requirements, location and travel feasibility, regional workload balance, and live job completion probability, accounting for the downstream impact on other appointments before any reallocation is proposed. It surfaces ranked options with the reasoning behind each, so dispatchers have the context to make a fast, informed decision. #### Appointment Renegotiation & Customer Communication Manages customer-facing interactions to reschedule appointments, propose alternatives, and maintain clear, consistent communication throughout the process. This skill generates new appointment options ranked by SLA compliance, travel efficiency, and customer priority, then engages the customer via email, SMS, or Teams, handles back-and-forth responses, refines options based on customer feedback, and captures the confirmed appointment for schedule update. It operates autonomously for routine renegotiations and escalates to dispatchers when customer preferences or constraints require judgment. #### PSO Schedule Exception Explanation & Dispatcher Support Explains schedule changes and optimization decisions by translating PSO outputs into clear, actionable insights for dispatchers. This skill identifies the root cause of an exception, analyzes schedule inefficiencies and available slack capacity, and surfaces trade-off options with clear impact explanations, helping dispatchers understand not just what changed, but why, and what the consequences of each resolution option would be. This is particularly valuable for exceptions where PSO has produced a recommendation that requires dispatcher confidence before it can be actioned. ## Dispatch Management Workflow The following table outlines the full exception management workflow from initial detection through to stakeholder notification. | **Stage** | **What the System Does** | **Action / Output** | **Channel** | | ---------------------------------- | ------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------- | ------------------- | | **Exception Detection** | Detects live disruptions, delays, cancellations, SLA risks, parts readiness issues | Triggers exception workflow, identifies affected job, engineer, and schedule window | FSM / PSO / Teams | | **Context & Impact Analysis** | Evaluates schedule, SLA risk, engineer availability, and constraints; assesses downstream impact | Surfaces SLA breach risk, flags high-priority customers, identifies affected downstream appointments | FSM / PSO | | **Slot Generation & Optimization** | Creates best-fit appointment options ranked by SLA exposure, travel efficiency, and priority | Proposes engineer swap, resequencing, or new appointment slots; accounts for real-time availability and constraints | FSM / PSO | | **Customer Negotiation** | Engages customer via email, SMS, or Teams; proposes options; handles responses | Presents appointment options, iterates based on customer feedback, captures confirmed slot | Email / SMS / Teams | | **Confirmation & Schedule Update** | Confirms agreed slot, updates IFS Cloud schedule, reassigns engineer if needed | Schedule updated, engineer notified with job pack and relevant resources | IFS Cloud / PSO | | **Stakeholder Notification** | Notifies customer, engineer, and dispatch; logs the full decision | Confirmation sent, outcome recorded, full audit trail maintained | Email / SMS / Teams | **Exception paths:** When no suitable slot is available, the system expands the search window and escalates to the dispatcher with trade-off options. When a decision is high-impact or low-confidence, the exception is routed to the dispatcher with ranked scenarios and clear reasoning. When a customer rejects proposed options, the system iterates with new proposals or adjusts based on stated preferences before escalating if needed. ## Integration Points **IFS Cloud / PSO:** Core integration for live schedule data, engineer availability, SLA commitments, and appointment updates. The Dispatcher Assistant runs alongside PSO, reading optimization outputs and writing confirmed schedule changes back into the system. **ERP / Inventory Systems:** Provides parts availability context relevant to job feasibility and appointment timing. **CRM / Ticketing Systems:** Provides customer history, priority context, and communication records. **Email / SMS:** Primary channels for customer appointment renegotiation and confirmation. **Collaboration Platforms:** Microsoft Teams and email for human-in-the-loop escalation, dispatcher decision support, and engineer notification. **Knowledge Platforms:** Connects to knowledge sources to provide relevant job resources to newly assigned engineers. ### Human-in-the-Loop Design The Dispatcher Assistant is designed so that routine exception handling proceeds automatically and dispatcher judgment is applied to the decisions that genuinely require it. The system proceeds automatically when: * The exception is low-risk and a clear resolution exists within configured parameters * Customer communication follows a standard renegotiation pattern * Engineer reallocation meets all feasibility criteria without conflict * The action falls within approved thresholds for autonomous execution Dispatcher approval is required when: * The exception is high-impact or the confidence in the recommended action is low * No suitable slot is available within the standard search window * Trade-offs between SLA protection, customer impact, and engineer workload require judgment * The customer rejects proposed options and the situation requires a strategic decision When escalating to a dispatcher, the system surfaces ranked options with clear reasoning, impact explanations, and trade-off context, everything needed to make a confident decision in seconds rather than minutes. Once the dispatcher confirms, the system executes immediately and handles all downstream communication and system updates. ## Feedback and Continuous Improvement Dispatchers can rate the Dispatcher Assistant's recommendations with a thumbs up or thumbs down directly in the Teams interface. This feedback loop reinforces effective behavior and surfaces patterns for improvement over time, both in how exceptions are classified and in how reallocation options are ranked and presented. All exceptions, recommendations, decisions, and outcomes are logged with full traceability, providing the data needed for ongoing performance analysis, SLA auditing, and operational improvement. # Real-world scenarios ## Overview This article walks through one of the most common field service disruptions, a technician needing to leave early, and shows how the Dispatcher Assistant handles it end to end. The scenario is shown twice: once through the Microsoft Teams interface, and once through the Dispatch interface. Both result in the same outcome, illustrating that the Dispatcher Assistant operates consistently across whichever interface the dispatcher is working in. ### Scenario: Technician Leaves Early, Reassignment via Microsoft Teams A technician has to go home early, leaving a job unassigned for the remainder of the day. Without the Dispatcher Assistant, this would require the dispatcher to identify the affected job, check who has capacity, evaluate their suitability, make the reassignment, notify the new technician, and contact the customer, multiple steps across multiple systems, typically taking 10 to 20 minutes per disrupted job. #### Step 1: Exception Detected The Dispatcher Assistant detects that the technician needs to leave early and proactively surfaces the issue in Microsoft Teams before the dispatcher has spotted it manually. It immediately pulls the relevant job details, what was scheduled, the timing, and the expected duration, so the dispatcher has full context from the first message. #### Step 2: Available Resources Identified The system analyzes engineer availability and evaluates who has the capacity, skills, and travel feasibility to take the job. Two available technicians are surfaced as options. Alongside each option, the Dispatcher Assistant includes additional context, previous experience with similar jobs, current availability, location, and any other relevant factors, giving the dispatcher everything needed to make the right call quickly. #### Step 3: Dispatcher Makes the Selection The dispatcher reviews the options in Teams and selects the technician they want to assign. The decision takes seconds rather than the time it would have taken to manually investigate each option. #### Step 4: Reassignment Executed The Dispatcher Assistant populates the reassignment details and pulls any relevant resources for the job, documentation, job context, or other materials, sending them directly to the newly assigned technician. The schedule is updated automatically. #### Step 5: Feedback Captured The dispatcher gives a thumbs up or thumbs down on the recommendation. This simple feedback loop helps the Dispatcher Assistant track and improve its performance over time, both in how it ranks options and how it presents context for decision-making. #### Step 6: Customer Notified Automatically The Dispatcher Assistant emails the customer to let them know their schedule has changed. No manual email drafting, no risk of the notification being missed or delayed. The customer is informed promptly with a clear, professional message. **Result:** A process that would normally span multiple steps, multiple people, and significant back-and-forth is resolved end to end with a few keystrokes. The job is reassigned, the new technician is briefed, and the customer is notified, all handled through a single Teams interaction. ### Same Scenario: Reassignment via the Dispatch Interface The same exception, a technician leaving early, can also be handled directly through the Dispatch interface rather than Teams. The process and outcome are identical; the interface is different. The Dispatcher Assistant surfaces the notification in the Dispatch interface, providing the same level of detail: which job is affected, the original schedule, and how long it was set to last. Two available resources are presented with additional context to support the decision. The dispatcher selects the preferred technician, the job is reassigned, relevant resources are shared with the new technician, and the customer is automatically notified. The feedback mechanism works the same way, the dispatcher gives a thumbs up or down, and the full decision is logged with an audit trail. ## What This Scenario Illustrates **Exceptions are caught proactively, not reactively.** The Dispatcher Assistant surfaces the issue before the dispatcher has had to spot it themselves, early enough that a clean reassignment is straightforward rather than a last-minute scramble. **Options come with reasoning, not just names.** The system doesn't just show who is available, it provides the context behind each option, so the dispatcher's decision is informed rather than instinctive. Experience, availability, and location are all surfaced alongside the recommendation. **The dispatcher makes the decision; the system does the rest.** Once the dispatcher selects a technician, everything else, the reassignment, the technician briefing, the customer notification, happens automatically. The dispatcher's role is the judgment call, not the coordination work surrounding it. **Customer communication is never a separate step.** Notifying the customer is part of the same automated process, not an afterthought. The customer receives a timely, professional update without the dispatcher having to draft or send anything manually. **The interface is flexible; the process is consistent.** Whether the dispatcher is working in Teams or in the Dispatch interface, the same structured process applies, the same context, the same ranked options, the same automated execution after confirmation. This consistency reduces reliance on individual dispatcher experience and ensures exceptions are handled the same way across teams and regions. **Feedback improves future recommendations.** The thumbs up/down mechanism is not just a nicety, it creates a structured signal that helps the Dispatcher Assistant refine how it ranks and presents options over time. # FAQ ### Capabilities & functionality **Q: What does the Supplier Invoicing Manager do?** A: The Supplier Invoicing Manager is a Digital Worker that automates supplier invoice intake, validation, and exception routing for accounts payable teams. It ingests invoices received by email, parses and extracts the invoice data, validates each invoice against the corresponding purchase order and goods receipt, and either auto-approves the invoice or routes it for human review based on the outcome of that validation. ERP records are updated automatically once an invoice is approved. **Q: What is three-way matching and how does the Digital Worker perform it?** A: Three-way matching is the process of comparing an invoice against both the originating purchase order and the goods receipt to confirm that what was invoiced matches what was ordered and received. The Supplier Invoicing Manager executes this comparison automatically for every invoice it processes. It applies configurable tolerance logic to determine whether any differences fall within acceptable thresholds or require escalation. **Q: How does the Digital Worker handle invoices that do not match the PO or receipt?** A: When the Digital Worker detects a discrepancy that falls outside the defined tolerance thresholds, it flags the mismatch and routes the invoice to an AP team member for review. The reviewer receives the relevant details and context needed to assess the situation. Once they provide a decision, the Digital Worker takes follow-up action, including emailing the supplier to request a correction if needed. **Q: What happens after a valid invoice is approved?** A: Valid invoices are auto-approved without requiring manual touchpoints. The ERP is updated immediately to reflect the approval, and a copy of the invoice is attached to the record. The entire process runs without human intervention for invoices that match within tolerance. **Q: Can the Digital Worker handle service charges and line-item variances?** A: Yes. Charge and variance interpretation is one of the Digital Worker's core capabilities. It classifies variances across line items and determines whether each can be resolved within configured thresholds or requires escalation for review. This prevents unclear charges from stalling the approval process or slipping through unchecked. ### Control & Collaboration **Q: Does a human review every invoice?** A: No. Invoices that match within configured tolerance thresholds are auto-approved without any manual review. Human involvement is reserved for genuine exceptions, specifically invoices where discrepancies, unclear charges, or variances cannot be resolved automatically. This keeps your AP team focused on decisions that require judgment rather than routine processing. **Q: How does the Digital Worker involve AP team members when an exception occurs?** A: When an invoice requires human review, the Digital Worker routes it to the appropriate team member with the relevant details surfaced so they can act quickly and with full context. The reviewer assesses the situation and provides their feedback, after which the Digital Worker handles any required follow-up. **Q: What information does the reviewer receive when an invoice is escalated?** A: Escalated invoices are accompanied by the details needed to understand the discrepancy, including the invoice data, the relevant purchase order and receipt records, and the nature of the mismatch or variance. Reviewers can make informed decisions without having to manually locate supporting documents. **Q: How does the Digital Worker communicate with suppliers?** A: After a reviewer assesses a discrepancy and provides feedback, the Digital Worker sends an email directly to the supplier to flag the issue and request a correction. This removes the need for manual inbox monitoring or back-and-forth coordination by the AP team. ### Integration & Technology **Q: What systems does the Supplier Invoicing Manager integrate with?** A: The Supplier Invoicing Manager integrates with email platforms, enterprise applications including ERP and CRM systems, collaboration tools, and document platforms. This allows it to receive invoices from existing inbound channels and update records in the systems your team already uses. **Q: How does the Digital Worker receive invoices?** A: Invoices are captured from inbound email channels. The Digital Worker monitors these channels, ingests invoices as they arrive, and prepares them for processing automatically. No manual upload or intervention is required to initiate the process. **Q: How does it update the ERP after processing?** A: Once an invoice is approved, the Supplier Invoicing Manager updates the invoice status in the ERP directly. This keeps invoice progress and outcomes visible in your system of record without requiring manual data entry by the AP team. **Q: What collaboration and document platforms are supported?** A: The Digital Worker supports integration with collaboration platforms including Slack and Microsoft Teams, and document platforms including IFS and Google Drive. Refer to your implementation documentation for the full list of supported connectors available in your environment. ### Security & Reliability **Q: How is audit traceability maintained?** A: Every decision the Digital Worker makes is traceable. Discrepancies, the context in which they were detected, and how they were resolved are all recorded with full traceability. This means AP teams and auditors can review the rationale behind any approval or escalation after the fact. **Q: What prevents duplicate payments or overpayment?** A: The Digital Worker enforces consistent matching rules and applies idempotency controls to prevent invoices from being processed more than once. Combined with three-way matching against PO and receipt records, this significantly reduces the risk of duplicate payments and overpayment. **Q: How are exceptions handled consistently?** A: Exception routing follows defined tolerance and matching rules configured for your environment. Every invoice is evaluated against the same criteria, regardless of volume or who is processing the queue. Consistent rules also make it easier to audit exception decisions and identify patterns over time. ### Customization & Performance **Q: Can matching rules and tolerance thresholds be configured?** A: Yes. The Supplier Invoicing Manager uses configurable tolerance logic for three-way matching and auto-approval decisions. This allows organizations to define the thresholds that reflect their own procurement and finance policies, rather than applying fixed rules that may not suit their business. **Q: Can the Digital Worker be adjusted to route specific exception types differently?** A: Exception routing is governed by the rules configured in your environment. Invoices that fall outside tolerance thresholds are escalated with decision rationale, allowing your team to define how different types of discrepancies should be handled and by whom. **Q: What agentic skills does the Supplier Invoicing Manager use?** A: The Supplier Invoicing Manager is built on four agentic skills: Invoice Intake & Parsing, 3-Way Match, Charge & Variance Interpretation, and Exception Handling. Each skill handles a distinct part of the invoice processing lifecycle, and together they cover the full path from invoice receipt through to ERP update or supplier follow-up. # WIP - Case studies See how our customers are using this Digital Worker to transform their operations. # Inventory Replenisher Keeping inventory at the right levels is critical, but manual replenishment shouldn’t demand constant human effort. This knowledge base helps you get the most out of the Inventory Replenisher, covering setup, configuration, and efficient day‑to‑day replenishment management. ### Jump right in
Cover image
Watch & learnSee this Digital Worker in action./files/IniyzVQmRw3Net0b8IkX/pages/FfMRJgLoupC0rfHq2vih
Core conceptsEssential context for understanding how it all works./files/fZNxfxVnHxx6iNh5M9k9/pages/bevkSgmD2LjJRjWHaHby
FAQQuick answers to common questions./files/RBUXYX1hxsMuVshLZ5Ar/pages/6kabzv8C4TSqFryUzuNK
How-to guidesPractical, step-by-step guidance across a range of tasks/files/uC1RptEfw6kewkOX3sMG/pages/2Qs8NANzzJTvoXvABd7B
# Watch & learn Watch the video below to see this Digital Worker in action {% embed url="" %} # Core concepts Explore core concepts to build a solid understanding of the Digital Worker. These articles will give you the context needed to get the most out of your Digital Worker.
Cover image
FundamentalsLearn the full scope of what's possible/files/PKWzoCypd9fEWl1gt5o8/pages/B0I0w7s4RAzIaWB1NIam
Architecture & capabilitiesDiscover the components and integrations/files/MgisEby9ZvW43Wybsymq/pages/IlUwrI6EXD6SSOMwsQWt
Real-world scenariosPractical examples of the platform in action across common business use cases./files/I8b6oW47o0Zv7grfRIxS/pages/TH5HfuJ1RbbG4ehaGxWk
# Fundamentals ## Overview The Inventory Replenisher is an AI-powered Digital Worker designed to free inventory planners from the repetitive work of processing purchase requisitions so they can focus on the strategic supply management that actually requires human expertise. It continuously monitors ERP systems, validates supplier data, applies configured business rules, and automatically converts purchase requisitions into purchase orders, ensuring material availability while minimizing carrying costs and stockouts. By transforming manual, reactive replenishment into an intelligent, autonomous process, the Inventory Replenisher gives planners time back for strategic sourcing, supplier relationship building, demand planning, and category management. This is work that requires human judgment, negotiation skills, and strategic thinking. ## Key Benefits **Automated requisition processing:** Purchase requisition lines in planned status are evaluated, validated, and converted to purchase orders automatically when all configured conditions are met, without any manual review required for clean lines. **Consistent rule enforcement:** Business rules for supplier validity, minimum order quantities, price thresholds, and freight grouping are applied the same way every time, eliminating the inconsistency that comes from different planners making different judgment calls on routine processing. **Proactive exception visibility:** Lines that don't meet processing conditions are flagged immediately with specific reasons, giving planners clear visibility into what needs attention and why, rather than discovering problems downstream. **Freight load optimization:** When multiple requisitions exist for the same supplier, the system intelligently groups orders to minimize partial shipments and reduce transportation costs automatically. **Continuous background operation:** The Digital Worker runs on a configured schedule, hourly, daily, weekly, or event-triggered, processing requisition lines in the background without requiring planners to initiate or monitor the run. ## The Challenge It Solves Inventory replenishment is often reactive and manual. Planners must validate suppliers, apply supplier-specific rules, and convert requisitions into purchase orders across systems, a process that requires checking each line individually against multiple conditions before anything can be actioned. Disconnected systems and inconsistent rule application lead to delays, supply risks, excess inventory, and limited visibility into what has been processed versus what is still pending. In practice, this means planners are spending time on tasks like retrieving planned PR lines from IFS Cloud, reviewing each line against supplier rules and MOQ requirements, determining which lines can be actioned, manually converting valid PRs to POs, updating the ERP, and managing exceptions, all work that is repetitive, rules-based, and consumes time that could be spent on strategic decisions. ## Before and After ### Before Before the Inventory Replenisher, processing a batch of purchase requisition lines requires manual effort at every step:
{% stepper %} {% step %} Retrieve PR lines from IFS Cloud {% endstep %} {% step %} Review each PR line individually {% endstep %} {% step %} Review supplier rules on a per-line basis {% endstep %} {% step %} Determine supplier guidelines and applicability {% endstep %} {% step %} Convert valid PRs to purchase orders {% endstep %} {% step %} Update the ERP {% endstep %} {% step %} Manage exceptions separately {% endstep %} {% endstepper %} This process is reactive, prone to inconsistency, and typically takes 30 to 45 minutes per processing run, time that compounds across daily or weekly replenishment cycles. ### After With the Inventory Replenisher, the same cycle runs in three automated steps: {% stepper %} {% step %} Retrieve and review PR lines automatically {% endstep %} {% step %} Apply supplier rules and business conditions {% endstep %} {% step %} Create purchase orders for valid lines and route exceptions to planners {% endstep %} {% endstepper %} Valid lines are processed without any human involvement. Invalid lines are flagged with clear reasons. The entire run that previously took 30 to 45 minutes happens in under a minute. ## How It Works The Inventory Replenisher executes a five-step automated workflow that handles the complete replenishment cycle from requisition identification through purchase order creation. **Requisition Line Identification:** The Digital Worker queries all purchase requisition lines in planned status within a defined timeframe, typically the last 24 hours. It retrieves planned requisitions based on latest order date, filters by configured date range, applies processing limits (default 20 lines, configurable to thousands), and extracts key attributes including supplier information, part numbers, quantities, prices, and delivery requirements. **Supplier Validation and Grouping:** Before processing any orders, the system confirms each supplier is active and approved, validates that suppliers are properly connected to the relevant company and site, organizes requisition lines by supplier number to enable consolidated processing, and verifies that expected delivery timelines are available and reasonable. **Condition Evaluation and Business Rules:** The system applies customizable business rules to determine which lines should be released. These include supplier-specific validation criteria, minimum order quantity requirements, price thresholds to validate that line totals or order values fall within acceptable ranges, supplier exclusions to automatically reject requisitions from blocked suppliers, and freight load considerations to optimize order groupings and reduce transportation costs. **Release and Purchase Order Creation:** Valid lines are automatically released from planned to released status in IFS Cloud, and complete purchase orders are generated. Invalid lines are flagged with specific reasons logged. Exception notifications are sent via Teams, Slack, or email for any lines requiring manual review. A complete audit trail of all decisions, validations, and actions is maintained for compliance and analysis. **Continuous Monitoring:** The Digital Worker runs on a configured schedule, daily, hourly, weekly, or event-triggered, continuously processing in the background without requiring planner initiation. # Architecture & capabilities ## Overview The Inventory Replenisher is built on a set of integrated agentic skills coordinated by a Digital Worker agent. This article covers the core skills, agent modes and configuration options, the full replenishment workflow, integration points, and how the system determines when to process automatically versus when to route to a planner. ## Agentic Skills The Inventory Replenisher operates through five core skills, each handling a distinct part of the replenishment process. #### Actionize Planning Outcomes Translates planning recommendations, shortages, and demand signals into actionable replenishment steps rather than leaving them as static insights. This skill bridges the gap between what the planning system has identified as needed and what actually needs to happen in IFS Cloud, turning MRP signals, reorder point triggers, and project demand into executable replenishment actions. #### Supplier Validation Confirms that replenishment actions are tied to the correct supplier based on sourcing rules, contracts, and approved vendor lists. Before any purchase order is created, this skill verifies that the supplier is active and approved in the system, properly connected to the relevant company and site, and that expected delivery timelines are available and valid. #### Conditional Selection Evaluates business conditions such as MOQ, value thresholds, lead times, and approval rules to determine whether actions can be automated or require human review, applied per supplier, per site, or per any other configured dimension. This is the skill that enforces consistency: the same rules are applied to every line, every run, regardless of which planner would have otherwise handled it manually. #### Purchase Order Creation Generates purchase orders automatically when all conditions are met, ensuring accuracy in quantities, pricing, and delivery dates. This skill handles both the creation of new purchase orders and the decision of whether to add lines to an existing purchase order or create a new one, logic that can be customized through the Digital Worker's configuration. #### Stock Re-Allocation Optimizes inventory by reallocating available stock across sites or demands to prevent shortages while minimizing excess inventory. When stock exists elsewhere in the network that can address a shortage before triggering a new external procurement, this skill identifies and acts on that opportunity. ## Agent Modes and Configuration The Inventory Replenisher offers two agent configurations to match the complexity of different replenishment environments. **Basic Agent Mode** uses standard language models for straightforward requisition processing with clear business rules. This mode is suitable for environments with well-defined supplier conditions and predictable replenishment patterns. **Reasoning Mode** employs advanced reasoning models for complex scenarios requiring deeper analysis of business rules, supplier conditions, and freight optimization. This mode is appropriate when supplier-specific rules are numerous or interdependent, when freight grouping logic is complex, or when exception handling requires more nuanced evaluation. In Reasoning Mode, administrators can configure: * Reasoning effort: Controls the computational effort applied to rule evaluation and supplier selection * Verbosity: Adjusts the detail level in responses and logging for transparency and debugging * Concurrent processing: Enables parallelization of supplier processing to improve performance when handling multiple suppliers simultaneously ## Replenishment Workflow The following table outlines the full replenishment workflow from demand signal monitoring through to exception handling. | **Stage** | **What the System Does** | **Action / Output** | **Channel** | | ---------------------------- | ------------------------------------------------------------------------------------------------------------------ | -------------------------------------------------------------------------------------------------------------------------------------------------- | ----------- | | **Demand Signal Monitoring** | Monitors MRP, order points, and project demand; monitors purchase requisition lines in planned state | Initiates replenishment workflow | IFS | | **Replenishment Decision** | Applies business constraints including MOQ, lead time, and supplier availability | Decides whether to procure or fulfilll internally; routes to sourcing if needed | IFS | | **PR Creation & Enrichment** | Checks requisition with correct supplier, quantity, and constraints | Auto-checks PRs; enriches with context for faster approvals | IFS | | **PR to PO Conversion** | Evaluates PR readiness for conversion; aligns with ERP authorization rules; detects safe-to-auto-release scenarios | Auto-releases valid PRs; converts PR to PO; routes to Supplier Order Manager for execution | IFS | | **Exception Handling** | Identifies MOQ conflicts, supplier constraints, and abnormal demand; detects ambiguity in decision-making | Triggers human-in-the-loop workflows; suggests alternatives such as reallocation, delay, or order splitting; resumes automated flow after decision | Teams / IFS | ## Tools and Integration Layer The Inventory Replenisher uses a comprehensive set of tools to interact with IFS Cloud and connected systems: **Get Planned Requisition Lines:** retrieves all requisition lines in planned status within the specified date range **Validate Supplier:** confirms supplier status, company assignment, and approval conditions Check Business Rules: applies customizable conditions for MOQ, pricing thresholds, and supplier-specific requirements **Auto-Release Requisition:** changes the status of valid lines from planned to released **Create Purchase Order:** generates complete PO records in IFS Cloud with all validated information **Log Results:** stores detailed execution logs for audit and troubleshooting **Send Notifications:** alerts users through Teams, Slack, or email for exceptions and approvals ## Integration Points **ERP Platforms:** Connects directly with IFS Cloud and other enterprise systems to read requisition lines, validate supplier data, and write purchase orders automatically. **Email Systems:** Integrates with Microsoft Outlook and Gmail for supplier communications and exception notifications. **Collaboration Platforms:** Uses Microsoft Teams and Slack for exception notifications, approval workflows, and real-time visibility into processing status. **Planning Systems:** Integrates with material planning (MRP), demand forecasting modules, and automated planning systems to receive requisitions from multiple sources. **IoT Sensors:** Receives demand signals from connected equipment and sensors to inform replenishment decisions. **Logistics/TMS:** Connects with transportation management systems for delivery coordination and inbound visibility. **CRM Systems:** Connects with Salesforce, HubSpot, and Qoncierge to maintain supplier relationship context and historical performance data. ## Human-in-the-Loop Design The Inventory Replenisher is designed so that routine replenishment processing happens automatically and human judgment is applied to lines and scenarios that genuinely require it. The system proceeds independently when: * Suppliers are valid, active, and approved in the system * Requisition lines meet minimum order quantity requirements * Line totals fall within configured price thresholds * Orders align with freight load optimization rules Human review is required when: * Suppliers are invalid or blocked in the system * Requisitions do not meet minimum order quantities * Line totals exceed configured price thresholds * Forecast deviations or unexpected demand patterns are detected * High-value orders require strategic judgment When exceptions are detected, the system identifies the specific validation failures preventing automatic processing, sends a detailed notification via Teams or Slack with complete context, provides a clear reason for each invalid requisition line, and maintains a complete audit trail of all decisions and validations. ## Testing and Validation The platform includes an evaluation environment for testing and development. Teams can trigger processing runs manually with specific date ranges, monitor execution in real time as the agent processes each supplier, review validation results for each requisition line, and observe purchase order creation attempts. Debug mode provides detailed visibility into inputs and outputs of each validation step. Execution summaries show total valid lines processed, invalid lines flagged, purchase orders created, and any failures, with complete audit trails available in detailed logs. # Real-world scenarios ## Overview This article walks through the Inventory Replenisher in action, illustrating how it processes a batch of purchase requisition lines, validating each one against configured business rules, creating purchase orders for lines that meet all conditions, and flagging the rest for planner review. It also covers how the system can be triggered manually for visibility into its decision-making, and how it runs as a background worker in normal operation. ### Scenario: Automated Requisition Processing with Mixed Results This scenario covers a typical replenishment run where the Digital Worker processes a batch of purchase requisition lines, some of which meet all configured conditions and some of which do not. #### Setup: Configuring Business Rules Before the Digital Worker runs, administrators configure the business rules that govern which requisition lines can be processed automatically. These rules act as the control center for the replenishment process and can include: * Supplier validity requirements (only process lines where the supplier is active and approved) * Minimum order quantity conditions (lines must meet the supplier's MOQ to proceed) * Price threshold limits (line totals or order values must fall within acceptable ranges) * Supplier exclusions (automatically reject requisitions from blocked suppliers) * Freight load grouping logic (how lines should be grouped into purchase orders) These rules are configured once and applied consistently to every processing run. They can be updated at any time as business requirements change. #### Normal Operation: Background Processing In normal operation, the Inventory Replenisher runs on a configured schedule, every hour, every day, every week, or triggered by a specific event. It processes in the background continuously, retrieving planned requisition lines from IFS Cloud, applying the configured business rules, creating purchase orders for valid lines, and notifying planners of any exceptions, all without requiring any manual initiation. #### Manual Trigger for Visibility For teams who want to observe the Digital Worker's decision-making process in real time, such as during initial setup or validation, the system can also be triggered manually via Microsoft Teams. When invoked through Teams, the Digital Worker fetches the current planned requisition lines and returns a detailed report showing how each line was evaluated before any purchase orders are created. This gives planners full visibility into which lines matched which conditions and why, so they can verify the logic before instructing the system to proceed. #### Processing the Requisition Lines In this example, the Digital Worker is triggered and retrieves 10 purchase requisition lines currently in planned status. For each line, it pulls the part number, ordered quantity, minimum order quantity, supplier details, and pricing, and evaluates every line against the full set of configured business rules simultaneously. The results come back within seconds: * 3 lines are valid, each one meets the supplier validity condition, the minimum order quantity condition, and the price threshold condition. These lines are ready to be converted to a purchase order. * 7 lines are invalid, each one fails one or more conditions. The Digital Worker identifies exactly which condition each line failed and reports this clearly, so planners know precisely what needs to be resolved before those lines can be processed. #### Purchase Order Creation The planner reviews the results, confirms the valid lines, and instructs the Digital Worker to proceed. The system auto-releases the three valid lines from planned to released status in IFS Cloud and creates a single purchase order, PO 143, containing all three lines with correct quantities, pricing, and delivery details. The Digital Worker also applies grouping logic to determine whether these lines should be added to an existing purchase order or require a new one. In this case, a new purchase order is created. This grouping behavior is fully configurable through the Digital Worker's instructions. The purchase order is immediately visible in IFS Cloud, already created and released. The entire process, from retrieving the 10 lines to creating the purchase order, takes under a minute. #### Exception Handling The 7 invalid lines are not silently discarded. The Digital Worker sends a notification via Microsoft Teams with the details of each invalid line and the specific reason it could not be processed. Planners can then address each issue, whether that means updating supplier information, negotiating a minimum order quantity waiver, or deferring the line, and re-run the process once the issues are resolved. ### What This Scenario Illustrates **The system evaluates every line against every rule simultaneously.** Rather than a planner working through lines one by one, the Digital Worker assesses the full batch in seconds and classifies each line as valid or invalid with clear reasoning. **Business rules are enforced consistently.** The same conditions apply to every line in every run, regardless of volume or which team member would otherwise have handled it. There is no variation in how rules are interpreted. **Valid and invalid lines are handled separately.** Valid lines proceed to purchase order creation immediately. Invalid lines are flagged with specific reasons and held for planner review, the two paths are handled in parallel without one blocking the other. **Planners see the decisions before they're executed.** When using manual trigger mode, planners can review the full evaluation report before any purchase orders are created. This transparency gives teams confidence in the system's logic and makes it easy to identify configuration adjustments if needed. **Grouping logic reduces unnecessary purchase orders.** The system intelligently decides whether to add lines to an existing purchase order or create a new one, applying freight load optimization to consolidate where possible, reducing transportation costs and administrative overhead. **The background mode removes the process entirely from planner workload.** Once configured, the Digital Worker runs on schedule without requiring any planner initiation, monitoring, or follow-up for clean lines. Planners only engage when exceptions require their judgment. # FAQ ### Capabilities & functionality **Q: What does Inventory Replenisher do for me?** A: It automates stock management by monitoring inventory levels, validating supplier data, and triggering replenishment orders automatically. It ensures products are available when and where they're needed, without manual tracking or intervention. **Q: How does Inventory Replenisher decide when to create a purchase order?** A: It continuously compares on-hand quantities, reorder points, and forecasted demand. When stock drops below a defined threshold, it automatically generates and validates purchase requisitions. **Q: Can it work across multiple locations or warehouses?** A: Yes. Inventory Replenisher supports multi-site operations and can independently assess stock positions for each location to maintain balanced inventory across the network. **Q: Does it handle supplier and pricing validation?** A: Absolutely. It verifies each supplier's status, pricing terms, and lead times before generating orders, ensuring every replenishment aligns with approved vendor policies. **Q: Can it adapt to changes in demand patterns?** A: Yes. It continuously learns from actual consumption and can factor in seasonal trends or forecast updates, adjusting order timing and quantities dynamically. **Q: What happens if stock levels change while replenishment is in progress?** A: The digital worker recalculates requirements in real time before finalizing purchase orders, ensuring over-ordering or duplication doesn't occur. ### Control & collaboration **Q: How can users monitor or trigger replenishment runs?** A: Users can view progress and initiate runs directly through Microsoft Teams, Slack, or the ERP interface. Real-time updates show which items are replenished and which are awaiting approval. **Q: Can planners intervene during the replenishment cycle?** A: Yes. You can define control points where the system pauses for human approval — such as when order value, supplier, or delivery timelines fall outside normal parameters. **Q: What if I need to pause or reschedule automatic replenishment?** A: Runs can be paused, delayed, or manually restarted at any time. Scheduling parameters are configurable, allowing daily, weekly, or event-based triggers. **Q: Will Inventory Replenisher notify me of exceptions?** A: Yes. If validation fails — such as an inactive supplier, pricing mismatch, or unavailable lead time — the system sends notifications in Teams, Slack, or email for quick review and correction. **Q: How does it handle collaboration between teams?** A: It shares replenishment reports, alerts, and audit logs in real time, so planners, procurement, and warehouse teams stay aligned without needing manual follow-ups. **Q: Can managers approve orders directly within chat tools?** A: Yes. Approvals and overrides can be executed directly within Teams or Slack messages, streamlining communication while maintaining oversight. ### Integration & technology **Q: How does Inventory Replenisher connect to ERP systems?** A: It integrates via secure APIs with platforms like IFS Cloud to read requisition lines, validate supplier details, and create purchase orders directly in the system. **Q: Can it pull data from forecasting or planning modules?** A: Yes. It can ingest signals from demand planning, MRP, or AI forecasting tools to refine replenishment timing and quantities. **Q: Does it integrate with communication platforms?** A: Absolutely. Teams, Slack, Gmail, and Outlook integrations allow the digital worker to send status updates, request approvals, and provide summaries without leaving your workspace. **Q: How does it handle large-scale data processing?** A: It's built for scalability, capable of evaluating thousands of requisition lines per run, grouping them by supplier, and generating POs efficiently. **Q: What happens if the ERP system or API connection fails mid-run?** A: The process automatically queues pending actions and retries when the system becomes available, ensuring no data or transaction loss. **Q: Can it support hybrid ERP environments?** A: Yes. Inventory Replenisher can work across multiple ERP instances or versions, maintaining data consistency through a unified integration layer. ### Security & reliability **Q: How are replenishment actions tracked?** A: Every action — validation, exception, and purchase order creation — is logged in the ERP audit trail and tagged with a digital worker ID for traceability. **Q: Is sensitive supplier and inventory data secure?** A: Yes. Data remains within your ERP and communication environments. The digital worker communicates via encrypted APIs and never stores sensitive data externally. **Q: How does it ensure process reliability?** A: It includes built-in retry mechanisms, queue management, and exception alerts to maintain process continuity even during temporary disruptions. **Q: Can we audit its decisions or view a history of changes?** A: Yes. A full audit trail is available, detailing every decision, timestamp, and user interaction to support compliance and continuous improvement. **Q: What happens if a scheduled run encounters unexpected input data?** A: The digital worker isolates invalid lines, logs the issue, and continues processing valid ones, avoiding total run failure. **Q: How does it ensure consistency between automated and manual actions?** A: All automated and manual transactions share the same audit trail and data source, eliminating duplication or data mismatches. ### Customization & performance **Q: Can Inventory Replenisher be configured for our business rules?** A: Yes. You can define reorder points, approval thresholds, and supplier selection logic based on your company's unique policies. **Q: Can it handle industry-specific replenishment models?** A: Absolutely. It supports sector-specific approaches like make-to-stock, just-in-time, or vendor-managed inventory. **Q: How do we measure its impact?** A: Key performance indicators include replenishment accuracy, automation rate, time saved per cycle, and reduction in manual purchase order creation. **Q: How are adjustments to parameters managed?** A: Parameters such as reorder points, approval rules, and scheduling frequencies can be updated directly through configuration settings — no coding required. **Q: Can it evolve with changing business priorities?** A: Yes. Rules and thresholds can be continuously refined as your operations, demand patterns, or supplier landscape evolve. **Q: What results can teams expect after deployment?** A: Most organizations see faster order cycles, fewer stockouts, and measurable reductions in manual administrative time within weeks of activation. # WIP - Case studies See how our customers are using this Digital Worker to transform their operations. # Knowledge Manager Managing organizational knowledge takes diligence, but tracking and retrieving information shouldn't slow teams down. This knowledge base guides you through using the Knowledge Manager to organize and surface the right information in daily operations. ### Jump right in
Cover image
Watch & learnSee this Digital Worker in action./files/IniyzVQmRw3Net0b8IkX/pages/RLHJHfNwO7uO9lkCp3M7
Core conceptsEssential context for understanding how it all works./files/fZNxfxVnHxx6iNh5M9k9/pages/6eA8J3ah3MiGbBfK6KrV
FAQQuick answers to common questions./files/RBUXYX1hxsMuVshLZ5Ar/pages/XSXnuOddnhLS61zCi0v1
How-to guidesPractical, step-by-step guidance across a range of tasks/files/uC1RptEfw6kewkOX3sMG/pages/2Qs8NANzzJTvoXvABd7B
# Watch & learn Watch the video below to see this Digital Worker in action {% embed url="" %} # Core concepts Explore core concepts to build a solid understanding of the Digital Worker. These articles will give you the context needed to get the most out of your Digital Worker.
Cover image
FundamentalsLearn the full scope of what's possible/files/PKWzoCypd9fEWl1gt5o8/pages/kwfmgrQeoVKpESumzcJB
Architecture & capabilitiesDiscover the components and integrations/files/MgisEby9ZvW43Wybsymq/pages/NTAMda6AOQvC3lTvPOff
Real-world scenariosPractical examples of the platform in action across common business use cases./files/I8b6oW47o0Zv7grfRIxS/pages/TA6DGNDYgeETKlYFxb82
# Fundamentals ## Overview The Knowledge Manager is an AI-powered Digital Worker designed to put trusted, relevant answers directly in the flow of work for service and operations teams. It interprets questions in natural language, searches across every connected knowledge source simultaneously, and returns context-specific guidance with citations, so teams can act quickly and with confidence rather than spending time searching, second-guessing, or escalating unnecessarily. Beyond retrieval, the Knowledge Manager continuously monitors interactions to identify missing, low-confidence, or conflicting knowledge. Where gaps are detected, it flags them automatically and supports structured content creation or enrichment, so the knowledge base evolves based on actual field demand rather than static documentation cycles. ## Key Benefits **Answers at the point of need:** Guidance is delivered directly in Microsoft Teams or other configured interfaces, no system switching, no manual document browsing, no searching across multiple tools. **Context-specific, not generic:** Responses are tailored to the specific asset, configuration, and operational scenario rather than returning broad documentation that may or may not apply. **Trusted results with citations:** Every response includes source links so users know exactly where the guidance comes from and can assess its authority. Confidence indicators help users understand how reliable the answer is. **Continuous knowledge improvement:** Every interaction generates signals about what knowledge is working, what is missing, and what needs review. Gaps are flagged automatically and draft knowledge can be generated for human approval, ensuring the knowledge base improves over time. **Consistent outcomes across teams:** All users receive guidance drawn from the same validated sources, reducing the inconsistency that comes from relying on individual expertise or informal knowledge sharing. **Governed by design:** Role-based visibility, escalation paths, and full auditability ensure that human oversight is maintained where it matters, particularly in high-compliance or safety-sensitive environments. ## The Challenge It Solves Accessing knowledge across service operations is fragmented, inconsistent, and time-consuming. Engineers and agents typically search across multiple systems, manuals, SharePoint, ERP records, ticket histories, PDFs, often relying on incomplete or outdated information. Critical knowledge gaps go unidentified, leading to repeated issues, slower resolutions, and increased operational risk. In practice, this looks like engineers spending 15 to 20 minutes per job on information lookup, searching 4 to 6 systems before finding a relevant answer, encountering conflicting guidance that applies to slightly different configurations, escalating to senior colleagues when documentation is unclear, and applying best-guess resolutions that may not fully resolve the issue. Knowledge from resolved jobs is rarely captured in a reusable form, so the same questions surface repeatedly. ## Before and After ### Before Before the Knowledge Manager, resolving a field issue involves manual searching at every stage: {% stepper %} {% step %} Search across multiple systems, manuals, SharePoint, tickets, emails {% endstep %} {% step %} Interpret incomplete or unclear problem context {% endstep %} {% step %} Review multiple documents and sources {% endstep %} {% step %} Compare conflicting or generic guidance {% endstep %} {% step %} Ask colleagues or escalate for support {% endstep %} {% step %} Apply a best-guess resolution {% endstep %} {% step %} Leave knowledge uncaptured or buried in free-text notes {% endstep %} {% endstepper %} The result is high search effort, inconsistent answers, slow resolution times, and knowledge that never gets reused. ### After With the Knowledge Manager, the same process runs in four steps: {% stepper %} {% step %} Interpret the query, context, and asset configuration automatically {% endstep %} {% step %} Retrieve, validate, and rank knowledge across all sources simultaneously {% endstep %} {% step %} Deliver precise guidance and surface cited sources in a single response {% endstep %} {% step %} Capture feedback and flag gaps for continuous improvement {% endstep %} {% endstepper %} ## How It Works **When a question comes in:** the Knowledge Manager forms a working hypothesis before retrieving any information, interpreting the intent behind the query, identifying the asset type and configuration, assessing urgency and operational risk, and determining which knowledge sources are most relevant. It then searches all connected sources simultaneously and returns ranked results with confidence signals and citations. **If the query lacks sufficient context for a high-confidence response:** the system asks a targeted follow-up question rather than returning generic guidance. When the right answer exists, it comes back as specific, actionable steps, not a document dump. **Every interaction is tracked:** confidence levels, follow-up questions, resolution outcomes, and recurring patterns are all monitored in the background. This continuous monitoring powers the knowledge gap detection capability, ensuring that what engineers keep asking but can't find reliable answers to is surfaced for content review and enrichment. **When a job is resolved:** the Knowledge Manager can convert the interaction and outcome into structured knowledge, drafted for human review and, once approved, published back into the knowledge base for future use. ## Knowledge Orchestration The Knowledge Manager operates across five continuous functions: **Seek:** A question is raised from the field or support desk. The Knowledge Manager captures context at source, asset ID, configuration, fault pattern, location, user role, risk level, and enriches the query before retrieval begins. **Surface:** Unified search runs across structured and unstructured sources simultaneously, manuals, historical service logs, prior resolutions, bulletins, tickets, and safety updates. Results are ranked by reliability and contextual match. Generic or unsafe guidance is suppressed. **Monitor:** The system continuously watches for changes that invalidate, improve, or create knowledge, source updates such as manual revisions and safety notices, changes in circumstances such as asset upgrades and firmware changes, and field outcomes such as confirmed fixes, rejected workarounds, and repeat visits. **Strengthen:** Gaps, conflicts, and low-confidence guidance are detected automatically. The system highlights the difference between what teams keep asking and what the knowledge base can actually answer, flagging missing configuration steps, recurring searches, and unreliable sources for review. **Generate:** Where knowledge is missing or repeatedly requested, the Knowledge Manager supports structured creation, drafting service log updates, proposing new content, generating configuration-specific guidance, routing everything through a human approval process before it is published back into the knowledge base. # Architecture & capabilities ## Overview The Knowledge Manager is built on five agentic skills that work together to deliver contextual knowledge retrieval and continuous knowledge improvement. This article covers the core skills, how the system is deployed and accessed, integration points, and how human oversight is maintained throughout. ## Agentic Skills #### Conversational Query Interpretation Understands user intent and context from natural language queries, identifying intent, urgency, asset type, and operational risk without relying on keyword matching. This skill forms a working hypothesis before any retrieval begins, enriching the query with asset and service history metadata, distinguishing troubleshooting requests from procedural queries, and routing to the most appropriate knowledge sources. If the query lacks sufficient context for a high-confidence response, it asks a targeted follow-up question rather than proceeding with incomplete information. #### Multi-Source Knowledge Search Searches simultaneously across multiple structured and unstructured systems to gather relevant information, manuals, engineering documentation, bulletins, historical work orders, case resolutions, similar asset failures, and prior repair outcomes. Unlike basic search, this skill cross-checks content consistency across sources, weighs recency and outcome history, and assembles a unified operational view rather than returning individual documents. #### Context-Aware Retrieval Fetches knowledge tailored to the specific asset, configuration, and operational scenario. Using asset ID, configuration variant, service history, and job context, this skill delivers configuration-specific troubleshooting steps and suppresses non-applicable or outdated instructions. The result is guidance that applies to the exact situation the user is facing, not generic documentation that requires interpretation. #### Relevance Ranking Evaluates and prioritizes validated, configuration-aligned guidance while suppressing outdated or low-confidence content. This skill ranks results based on source validation, usage signals, and resolution effectiveness across historical cases, returning guidance with clear confidence indicators so users understand how reliable each piece of guidance is. High-confidence, validated steps are surfaced first; unsafe or generic content is filtered out. #### Knowledge Gap Detection Identifies repeated queries, workarounds, and unresolved patterns to surface missing or weak documentation for continuous improvement. This skill monitors confidence levels and outcome signals from every interaction, detects recurring questions that the knowledge base cannot answer well, and automatically flags gaps for review. Where gaps are identified, it supports structured content creation, generating draft knowledge articles from field feedback and routing them through a human approval process before they are published. ## Deployment and Access The Knowledge Manager is a user-invoked conversational agent. It responds to questions raised directly by engineers, agents, or planners, it does not run as a background processing worker. **Primary interface:** Microsoft Teams, accessible on both desktop and mobile. Teams provides the conversational interface through which users submit questions, receive guidance, follow up, and rate responses. **Additional interfaces:** * API access for integration into existing customer chatbots or service platforms * Embedded chat widget deployable with a single line of code into any web interface * Deep integration into existing bots such as Salesforce Einstein, where the Knowledge Manager acts as a callable skill that receives context from the host bot and returns structured, governed responses This interface-agnostic design means the Knowledge Manager can enhance existing conversations and service platforms without requiring users to switch tools. ## Connected Knowledge Sources The Knowledge Manager connects to a wide range of knowledge sources across an enterprise: * **Manuals and technical documentation:** equipment manuals, engineering bulletins, safety notices * **Service management systems:** ServiceNow, Zendesk, Freshdesk, Salesforce Service Cloud * **Knowledge platforms:** Confluence, Notion, Guru, SharePoint, Google Drive * **ERP and field service systems:** IFS Cloud (FSM, PSO, service requests), Dynamics * **CRM systems:** Salesforce, HubSpot, Kustomer, Gainsight * **Collaboration and document tools:** Microsoft Teams, Google Drive, Poka * **Historical service records and case notes:** resolved tickets, prior repair outcomes, field job notes Sources are configured based on what is relevant and authorized for each deployment. The system can operate across any combination of structured and unstructured sources simultaneously. ## Integration Points **Collaboration Platforms:** Microsoft Teams and Slack serve as the primary conversational interfaces for field engineers and support teams. **Enterprise Applications:** IFS Cloud, Salesforce, and other enterprise platforms provide asset data, service history, and operational context that enriches retrieval. **Support Systems:** ServiceNow, Zendesk, Freshdesk, and similar platforms provide ticket history and case resolution data. **Knowledge Platforms:** SharePoint, Confluence, Notion, Guru, and Google Drive provide document-based knowledge content. **Email:** Outlook and Gmail support communication workflows where needed. **Documents:** IFS Loops and Google Drive support document retrieval and knowledge base management. ### Human-in-the-Loop Design The Knowledge Manager is designed with human oversight built in at the points where it matters most, particularly in high-compliance and safety-sensitive environments. Knowledge retrieval proceeds automatically when: * The query contains sufficient context for a high-confidence response * Matching knowledge exists across connected sources * The guidance is validated and configuration-appropriate Human input is required when: * The query lacks sufficient context and the system asks a clarifying question * A knowledge gap is detected and draft content is generated for review and approval * A response is rated negatively and the interaction is flagged for review * Knowledge from a resolved interaction is converted to a draft article requiring approval before publication Role-based visibility ensures users receive guidance appropriate to their role and access level. Escalation paths are maintained for cases that require senior review. Full auditability means every query, response, confidence rating, and outcome signal is tracked, supporting compliance requirements and continuous quality improvement. ### Feedback and Continuous Improvement Every response can be rated with a thumbs up or thumbs down directly in the interface. This feedback, combined with follow-up question patterns, resolution confirmation signals, and recurring gap detection, feeds continuously into improving both response quality and knowledge content. The feedback loop operates at two levels: improving how the system retrieves and ranks existing knowledge, and identifying what new or updated knowledge needs to be created. Both are monitored automatically, with human review triggered at the content approval stage rather than requiring manual monitoring of individual interactions.
# Real-world scenarios ## Overview This article walks through two practical scenarios that illustrate how the Knowledge Manager supports field engineers in the flow of work. The first follows a technician resolving a fault using the Knowledge Manager through Microsoft Teams, covering the full interaction from initial question through to job closure and knowledge capture. The second illustrates what happens when a question cannot be fully answered, showing how the system handles gaps rather than returning unreliable guidance. ### Scenario 1: Field Fault Resolution via Teams This scenario follows a field service engineer encountering a fault code on site and using the Knowledge Manager to resolve it, without searching across multiple systems, calling a senior colleague, or leaving the issue partially resolved. #### Start of Day The engineer opens their job in the mobile app. Before they arrive on site, the Knowledge Manager has already assembled a contextual knowledge pack based on the asset configuration, service history, and any relevant manuals, so the engineer arrives with relevant context already available rather than starting from scratch on site. #### Fault Encountered A fault code appears that the engineer hasn't seen recently on this configuration. Rather than opening multiple systems to search for relevant procedures, the engineer submits a question in natural language through Microsoft Teams, describing the scenario they're dealing with. #### Knowledge Manager Responds The Knowledge Manager interprets the intent behind the question, identifies the asset type and configuration from the job context, and searches all connected knowledge sources simultaneously: manuals, historical service records, prior case resolutions, and engineering bulletins. Within seconds, a response comes back directly in Teams with: * The most relevant diagnostic steps for that specific asset configuration * Citations showing exactly which sources the guidance came from * Links to supporting documentation for reference The engineer does not need to open any other system. The guidance is specific to the asset they are working on, not a generic procedure that requires interpretation. #### Follow-Up and Clarification The engineer follows the diagnostic steps and adds some additional context about what they're observing. The Knowledge Manager responds with further detail, continuing the conversation naturally and maintaining the context of the original question throughout. Each response can be rated with a thumbs up or thumbs down. This feedback is captured automatically and contributes to improving future responses for similar queries. #### Issue Resolved The engineer follows the recommended steps and resolves the issue on the first visit. The interaction has replaced what would previously have been 15 to 20 minutes of searching across multiple systems, a phone call to a senior technician, and a potential follow-up visit due to incomplete or generic guidance. #### Job Notes Generated When the job is closed, the Knowledge Manager can generate a set of job notes from the interaction, structured from the conversation, the guidance used, and the outcome. The engineer reviews and submits them. These notes contribute to the knowledge base, ensuring that the resolution is captured in a reusable form rather than buried in free-text. ### Scenario 2: Handling a Knowledge Gap This scenario covers what happens when a question arrives that the Knowledge Manager cannot answer with high confidence, for example, a query about a non-standard configuration for which documentation is incomplete or missing. #### Question Submitted An engineer submits a question about safety or configuration steps for a non-standard installation. The Knowledge Manager interprets the query and begins searching connected sources. #### Low Confidence Detected The system finds some relevant content but identifies that the available guidance has low confidence for this specific configuration, perhaps because documentation is partial, conflicting, or has not been updated to reflect recent equipment changes. Rather than returning the low-confidence guidance as if it were reliable, the Knowledge Manager does two things: 1. Requests additional context from the engineer, asking a targeted follow-up question to narrow the scenario before returning guidance. This ensures that if a reliable answer exists, it can be retrieved more precisely. 2. Flags the gap in the background, registering that this configuration has been queried and that the available knowledge is insufficient to answer it with confidence. #### Guidance Returned with Appropriate Signals If a high-confidence answer can be assembled after clarification, it is returned with citations and confidence indicators. If the gap is genuine and cannot be resolved from current sources, the Knowledge Manager communicates this clearly rather than returning unreliable guidance, and escalation paths are available for cases that require senior review. #### Gap Triggers Content Action The flagged gap is logged automatically. If the same configuration is queried repeatedly without a satisfactory answer, the Knowledge Manager surfaces this pattern for review, identifying it as a knowledge gap that needs to be addressed. It can generate a draft knowledge article based on the interaction context, which is routed through a human approval process before being published back into the knowledge base. Over time, this means the knowledge base evolves based on what engineers are actually encountering in the field, not just what was documented when the system was first set up. ### What These Scenarios Illustrate **Knowledge is delivered in the flow of work.** Engineers don't leave Teams to find answers. The entire interaction, question, guidance, follow-up, job notes, happens in the same interface they're already using. **Guidance is specific, not generic.** The Knowledge Manager uses asset ID, configuration, service history, and job context to return steps that apply to the exact situation. Generic documentation that might not apply is filtered out. **Citations make trust explicit.** Every response includes sources, so engineers can see where the guidance comes from and assess its authority. Confidence indicators make the reliability of each answer visible. **The system asks rather than guesses.** When context is missing or confidence is low, the Knowledge Manager asks a targeted follow-up question rather than returning uncertain guidance. This is particularly important in safety-sensitive environments where acting on incorrect information carries real risk. **Gaps are captured, not ignored.** When knowledge doesn't exist or is insufficient, the system flags it automatically rather than letting the gap persist. Over time this drives continuous improvement of the knowledge base from real field demand. **Resolved jobs feed future answers.** Knowledge generated through interactions, job notes, resolution steps, workarounds, is structured and routed for review rather than left in free-text notes. This turns individual resolutions into reusable organizational knowledge. # FAQ ## Capabilities & Functionality **Q: What does the Knowledge Manager do?** A: The Knowledge Manager is a Digital Worker that gives service and operations teams fast access to trusted, relevant guidance in the flow of work. When a user submits a question, it interprets the intent behind it, searches across every connected knowledge source simultaneously, and returns ranked answers with citations and confidence indicators so users know exactly where the guidance comes from. It flags gaps in the knowledge base automatically, improves with user feedback, and can generate structured knowledge articles from resolved work for human review and approval. **Q: What sources does it search when answering a question?** A: The Knowledge Manager searches across all connected sources simultaneously, including manuals, PDFs, job histories, case notes, and resolved cases. It can connect to enterprise applications, support platforms, knowledge platforms, and document repositories. The specific sources available depend on the integrations configured for your environment. **Q: How does it handle ambiguous or incomplete questions?** A: When a question does not contain enough information to return a confident answer, the Digital Worker asks a targeted follow-up question to clarify what is needed. It interprets the intent behind queries, including asset context, urgency, configuration, and role, so it can ask the right question rather than returning low-confidence results or requiring the user to rephrase and resubmit. **Q: How does it rank and present answers?** A: Answers are ranked by relevance, confidence, and source authority. The Digital Worker surfaces the most appropriate guidance first and displays citations alongside each response so users can see exactly which source the guidance comes from. Confidence indicators give users a clear signal of how well-supported each answer is, which is particularly important in compliance-sensitive or safety-critical environments. **Q: Can it generate knowledge from resolved work?** A: Yes. After a complex issue is resolved, the Digital Worker can generate a structured draft knowledge article from the interaction. This draft is submitted for human review and approval before it becomes part of the searchable knowledge base. No unreviewed content enters the knowledge base automatically. ## Control & Collaboration **Q: Does a human review the guidance before it reaches the user?** A: Knowledge that already exists in the connected sources is surfaced directly to users with citations and confidence indicators rather than passing through a review step at query time. The human-in-the-loop model applies at the knowledge creation stage: articles generated from resolved work are reviewed and approved by a knowledge owner before they are added to the knowledge base and made available for future queries. **Q: How does the Digital Worker handle questions it cannot answer confidently?** A: When the Digital Worker cannot find a confident answer, it does not surface low-quality or speculative guidance. Instead, it flags the gap automatically for review by the knowledge team. The question pattern is tracked so that recurring unanswered questions become visible and can be prioritized for knowledge development. **Q: How does feedback from users improve the Digital Worker over time?** A: Each response can be rated with a thumbs up or thumbs down. This feedback is used to improve how the Digital Worker responds and which content it surfaces. The system also learns from confirmed outcomes and human review decisions, so the quality of guidance improves as the knowledge base and usage patterns develop. **Q: Can it generate job notes from an interaction?** A: Yes. At the end of an interaction, the Digital Worker can generate a set of job notes based on the conversation. These notes can be reviewed by the user and submitted automatically, reducing the time required for post-job documentation. ### Integration & Technology **Q: What systems and platforms does the Knowledge Manager integrate with?** A: The Knowledge Manager integrates with enterprise applications, support platforms, email, knowledge management platforms, collaboration tools, and document repositories. This allows it to search across the sources your teams already use and deliver answers through the channels where they already work. **Q: How do users access it?** A: Users access the Knowledge Manager through Microsoft Teams on desktop and mobile, as well as through other connected communication channels. The interface supports natural language questions and follow-up queries within the same conversation, so users can get further detail or run diagnostic steps without switching tools or starting over. **Q: What knowledge sources can it connect to?** A: The Knowledge Manager can connect to knowledge platforms including Notion, Confluence, and Google, support platforms including Zendesk and Intercom, enterprise applications including IFS and Salesforce, document repositories including Google Drive, and email platforms including Microsoft Outlook and Gmail. Refer to your implementation documentation for the specific connectors enabled in your environment. ### Security & Reliability **Q: How does the Digital Worker ensure the guidance it surfaces is trustworthy?** A: Every response includes citations that identify the source of the guidance and confidence indicators that reflect how well-supported the answer is. Users are never presented with guidance stripped of its provenance. This means teams can make informed decisions about how much weight to place on any given response, which is particularly important in high-compliance or safety-sensitive contexts. **Q: What prevents outdated or unsafe guidance from being surfaced?** A: The Relevance Ranking skill surfaces validated steps and suppresses outdated or unsafe guidance based on source authority and confidence scoring. Knowledge gaps and weak guidance are flagged for review rather than surfaced to users. The human-in-the-loop approval process for new knowledge articles also prevents unreviewed content from entering the knowledge base. **Q: How is role-based access and visibility managed?** A: The Knowledge Manager supports role-based visibility, ensuring that users see guidance appropriate to their role and context. Escalation paths and traceability are built into the system, keeping humans in the loop where compliance or operational risk requires it. **Q: Is there an audit trail of interactions and guidance provided?** A: Yes. The Digital Worker is governed by design, with full visibility and auditability across interactions. This supports compliance requirements and gives operations and knowledge teams the oversight they need to manage guidance quality and track how the system is being used. ### Customization & Performance **Q: Is it suitable for high-compliance or safety-sensitive environments?** A: Yes. The Knowledge Manager is specifically designed for service and operations teams in high-compliance or safety-sensitive environments. The combination of source citations, confidence indicators, role-based visibility, human-in-the-loop knowledge approval, and full auditability makes it appropriate for contexts where acting on incorrect or outdated guidance carries real operational or safety risk. **Q: What agentic skills does the Knowledge Manager use?** A: The Knowledge Manager is built on five agentic skills: Conversational Query Interpretation, Context-Aware Retrieval, Knowledge Gap Detection, Relevance Ranking, and Multi-Source Knowledge Search. Together these skills cover the full path from interpreting a user's question through to surfacing ranked, cited guidance and identifying where the knowledge base needs to be strengthened. **Q: Can it be configured for specific asset types, teams, or operational contexts?** A: Yes. The Context-Aware Retrieval skill delivers asset- and configuration-specific guidance based on the operational context of the query. The sources it searches, the knowledge it prioritizes, and the escalation paths it applies can all be configured to reflect your organization's specific asset types, team structures, and compliance requirements.
# WIP - Case studies See how our customers are using this Digital Worker to transform their operations. # Material Replenisher The Material Replenisher simplifies how technicians find parts and create material requisitions, removing unnecessary system complexity from field work. By automating part searches and requests in IFS, it helps technicians focus on repairs and planners focus on keeping materials available where they’re needed most. ### Jump right in
Cover image

:circle-play:

Watch & learnSee this Digital Worker in action./files/IniyzVQmRw3Net0b8IkX/pages/Y7ynT78ahP8Yb5yzIQS8

:head-side-gear:

Core conceptsEssential context for understanding how it all works./files/fZNxfxVnHxx6iNh5M9k9/pages/IoOmhk70VXjnEIa8zxMt

:circle-question:

FAQQuick answers to common questions./files/RBUXYX1hxsMuVshLZ5Ar/pages/CV0gFM5VJzwM4wxYNdzf

:glasses-round:

Case studiesLearn from real world examples of success./files/c1GhIydX1yr8ZcJL60g6/pages/XY9JqXKM8r8QVZo1W3tJ

:hammer:

How-to guidesPractical, step-by-step guidance across a range of tasks/files/uC1RptEfw6kewkOX3sMG/pages/2Qs8NANzzJTvoXvABd7B
# Watch & learn Watch the video below to see this Digital Worker in action {% embed url="" %} # Core concepts Explore core concepts to build a solid understanding of the Digital Worker. These articles will give you the context needed to get the most out of your Digital Worker.
Cover image
FundamentalsLearn the full scope of what's possible/files/PKWzoCypd9fEWl1gt5o8/pages/qwxKR1yf2Uyl5f140716
Architecture & capabilitiesDiscover the components and integrations/files/MgisEby9ZvW43Wybsymq/pages/PJ8T9ALUD5wrEnbHYv3D
Real-world scenariosPractical examples of the platform in action across common business use cases./files/I8b6oW47o0Zv7grfRIxS/pages/cHkozkCVDn1uDeX5bve8
# Fundamentals ## Overview The Material Replenisher is an AI-powered Digital Worker designed to free field technicians from administrative work so they can focus on the repair work they were hired to do. It streamlines the process of finding parts and creating maintenance material requisitions (MMRs) within the IFS platform, enabling technicians to search for parts using natural language, check inventory availability across connected warehouses, and automatically generate requisitions, all without navigating complex system interfaces. By transforming manual, fragmented part requests into an intelligent, automated process, the Material Replenisher gives technicians their time back for hands-on field work and gives planners time for supply chain optimization. This is work that requires human expertise, problem-solving skills, and equipment knowledge. ## Key Benefits The Material Replenisher delivers value by automating part request workflows across the full replenishment cycle: **Instant part intelligence:** Natural language queries return availability, location, and compatibility information in seconds, keeping technicians focused on equipment repairs, not system navigation. **Zero back-office churn:** Eliminates phone calls and emails between technicians and planners, standardizing processes across all locations and reducing delays. **Error prevention built-in:** Intelligent validation catches duplicate orders and incorrect reservations before they create problems, while historical intelligence suggests proven parts. **Reduced equipment downtime:** Faster part availability gets technicians back to productive work quickly, minimizing costly equipment idle time. **Human expertise elevated:** Skilled field technicians focus on hands-on repairs while planners optimize supply chains and improve inventory planning. ### The Challenge It Solves Material replenishment is often manual and fragmented. Technicians and planners must identify parts, validate availability, and raise requisitions across multiple systems, leading to delays in part fulfilllment, incorrect part selection, excess or missing inventory, and heavy dependency on planners for routine requests. When technicians need parts, they traditionally must: * Search through extensive part catalogs with complex numbering systems * Verify part availability across multiple warehouse locations * Check inventory at sites connected to their user profile * Create maintenance material requisitions with proper documentation * Link parts to specific work orders and work tasks through multiple system screens High request volume combined with site-by-site variations leads to duplicate orders, incorrect reservations, and constant back-and-forth between technicians, planners, and procurement teams. ## Before and After ### Before Before the Material Replenisher, the replenishment process requires significant manual effort at every step: {% stepper %} {% step %} Receive technician request for part {% endstep %} {% step %} Search parts catalogs manually {% endstep %} {% step %} Identify correct parts or alternatives {% endstep %} {% step %} Check inventory availability {% endstep %} {% step %} Validate part details {% endstep %} {% step %} Create requisitions (MMR/MPR) {% endstep %} {% step %} Update ERP and follow up on fulfilllment {% endstep %} {% endstepper %} This manual process introduces high error rates, delayed technician support, and heavy planner dependency. #### After With the Material Replenisher, the same process is streamlined to three automated steps: {% stepper %} {% step %} Receive request from technician and understand the context {% endstep %} {% step %} Search for part by categories and sub-categories {% endstep %} {% step %} Create requisitions (MMR/MPR) {% endstep %} {% endstepper %} Routine requests flow through automatically. Exceptions are routed to the right person with full context already attached. ## How It Works The Material Replenisher operates through an automated workflow that handles the complete part request and requisition cycle. **User Identification:** The Digital Worker identifies the technician making the request using their email ID, retrieving the corresponding user ID from IFS Cloud. This determines which warehouse sites are connected to the technician, associates the correct user with created requisitions, and ensures proper authorization and access controls. **Information Extraction:** The system analyzes natural language queries to extract key information: part description, functional unit ID for the specific equipment, unit type classification, and product family category. This transforms unstructured queries like "I need air filters for functional unit XYZ" into structured search parameters. **Intelligent Part Catalog Search:** The system performs targeted searches using multiple strategies: by unit type to find parts designed for that specific equipment, by product family to find relevant parts within a category, and by searching historical work tasks associated with the functional unit ID to identify parts previously used for that equipment. **Verification and Validation:** Before proceeding, the system verifies that parts have been extracted correctly. If no valid information can be extracted from the query, it requests clarification from the technician rather than proceeding with incomplete data. **Availability Check:** When technicians request to check part availability, the Digital Worker identifies all warehouse sites connected to their user profile, checks inventory levels at each connected site, and reports back with specific locations where the part is in stock, including site ID and warehouse location details. **Maintenance Material Requisition Creation:** When technicians request to reserve a part or create a new MMR, the system automatically generates a maintenance material requisition using the technician's user ID, site ID associated with the selected warehouse, work order information, confirmed part number, and work task details. All records are properly documented and linked to the appropriate work orders. # Architecture & capabilities ## Overview The Material Replenisher is built on a set of integrated agentic skills that work together to handle the complete part request and requisition cycle. This article covers the core capabilities of the Digital Worker, how its components are structured, how it connects to your existing systems, and how it determines when to proceed independently versus when to involve a human. ## Agentic Skills The Material Replenisher operates through five core skills. Each handles a distinct stage of the replenishment process. #### Request Interpretation Understands technician requests by extracting the part need, unit context, description, and urgency from the message or form submission. This skill transforms unstructured natural language, such as "I need air filters for functional unit XYZ", into structured search parameters the system can act on. If the request lacks sufficient information to proceed, the system asks a targeted clarifying question rather than making the technician start over. #### Part Identification & Validation Identifies the correct inventory part and validates it against unit data and historical usage to prevent incorrect or duplicate selections. The system searches by unit type, product family, and historical work task data associated with the specific equipment. When multiple candidate parts with similar names or part numbers are found, all options are presented to the technician for selection rather than the system guessing. #### Multi-Site Inventory Lookup Checks inventory availability across multiple sites, warehouses, and locations to find usable stock and match it with the technician's location. The Digital Worker identifies all warehouse sites connected to the technician's user profile and checks inventory levels at each one, returning specific site IDs and warehouse location details. Before reporting a part as unavailable, it conducts a global search across all locations to make sure no stock has been missed. #### Fulfillment Decision Logic Determines the optimal fulfilllment path based on stock availability. When stock exists at a connected site, the system creates an internal transfer. When stock is not available anywhere, it initiates a purchase requisition (MPR). This decision is made automatically for clear cases, with the full context passed to a planner when human judgment is needed. #### Context-Aware Escalation Requests human confirmation only when information is missing, ambiguous, or requires judgment before proceeding. Rather than failing silently or making assumptions, the system pauses at the right moments, presenting options, asking for confirmation, and routing exceptions to planners with complete supporting context for fast resolution. ## Component Structure The Material Replenisher is built using a combination of tools and sub-agents coordinated by a Supervisor Agent: * User identity resolution tools * Part catalog search capabilities with multiple search strategies * Warehouse connectivity checks * Historical work task query engines * MMR and MPR creation and management automation The Supervisor Agent receives technician requests, orchestrates the appropriate skills in sequence, and determines when to proceed automatically versus when to surface a decision to the technician or planner. ## Replenishment Workflow The following table outlines the full replenishment workflow, from initial request intake through to fulfilllment actions. | **Stage** | **What the System Does** | **Action / Output** | **Channel** | | ------------------------------------------ | ------------------------------------------------------------------------- | ------------------------------------------------------------------------- | --------------- | | **Request Intake & Context Understanding** | Captures technician request and understands job/work order context | Structures request data, triggers material lookup | FSM / IFS / MWO | | **Part Identification & Validation** | Matches descriptions to parts, validates part numbers, detects duplicates | Identifies correct part, reduces misidentification, ensures data accuracy | IFS | | **Alternative Part Suggestion** | Identifies substitutes or equivalents, evaluates compatibility | Suggests alternative parts, improves fulfilllment success | IFS | | **Inventory Check & Availability** | Checks stock across locations, identifies availability constraints | Confirms availability, recommends sourcing or transfer | IFS | | **Requisition Creation (MMR/MPR)** | Validates part and request data, prepares ERP inputs | Creates MMR or MPR, reduces manual workload | IFS | | **Exception Handling** | Detects unclear or missing data, identifies part conflicts | Triggers human-in-the-loop review, provides resolution options | Teams | ## Integration Points The Material Replenisher connects with your existing technology ecosystem across several categories. **Collaboration Platforms:** Integrates directly with Microsoft Teams and Slack through widget configuration, allowing technicians to interact using natural language messages. Technicians can mention the IFS bot and issue commands directly through the channel interface, no separate system login required. **ERP Platforms:** Connects directly with IFS Cloud FSM/ERP modules to access work orders, work tasks, part catalogs, and warehouse inventory, and to create maintenance material requisitions automatically. **Field Service Management:** Monitors mobile work orders, work tasks, IoT fault alerts, and customer calls/portals to capture part requests from multiple sources. **Email:** Integrates with Outlook and Gmail to support request intake and communication. **Enterprise Applications:** Connects with enterprise platforms including IFS, Salesforce, and others for cross-system data access. **Documents:** Integrates with IFS Loops and Google Drive for document-based context and knowledge base updates. **Communication Channels:** Updates work orders, creates MPRs/MMRs, notifies customers, and updates knowledge bases automatically across all integrated systems. ### Human-in-the-Loop Design The Material Replenisher is designed around the principle that humans make strategic decisions while the Digital Worker handles administrative execution. The system proceeds independently when: * The part description is clear and unambiguous * The part is available in connected warehouse inventory * The request includes complete information including functional unit ID and work order * The technician has valid credentials with proper warehouse site connections Human input is required when: * Requests are vague or incomplete and need clarification * Multiple candidate parts with similar names require technician selection * Parts are not available in inventory and a planner decision on alternatives is needed * Conflicts or exceptions require manual resolution When clarification is needed, the system asks targeted questions, presents multiple options when similar parts are found, conducts global searches before reporting unavailability, and routes complex cases to planners with complete supporting context already attached. ## Testing and Validation The platform includes a comprehensive evaluation environment accessible through the Loops platform, allowing users to: * Input test queries directly and observe the Digital Worker's response in real time * Use Plan Flow visualization to see execution steps as they happen * Enable debug mode for granular visibility into input/output tracking, step-by-step execution, search results, and API call details * Review historical records of all interactions with processing time metrics and visual usage graphs This environment supports ongoing testing, optimization, and validation before deploying changes to live workflows.
# Real-world scenarios ## Overview This article walks through two practical scenarios that illustrate how the Knowledge Manager supports field engineers in the flow of work. The first follows a technician resolving a fault using the Knowledge Manager through Microsoft Teams, covering the full interaction from initial question through to job closure and knowledge capture. The second illustrates what happens when a question cannot be fully answered, showing how the system handles gaps rather than returning unreliable guidance. ### Scenario 1: Field Fault Resolution via Teams This scenario follows a field service engineer encountering a fault code on site and using the Knowledge Manager to resolve it, without searching across multiple systems, calling a senior colleague, or leaving the issue partially resolved. #### Start of Day The engineer opens their job in the mobile app. Before they arrive on site, the Knowledge Manager has already assembled a contextual knowledge pack based on the asset configuration, service history, and any relevant manuals, so the engineer arrives with relevant context already available rather than starting from scratch on site. #### Fault Encountered A fault code appears that the engineer hasn't seen recently on this configuration. Rather than opening multiple systems to search for relevant procedures, the engineer submits a question in natural language through Microsoft Teams, describing the scenario they're dealing with. #### Knowledge Manager Responds The Knowledge Manager interprets the intent behind the question, identifies the asset type and configuration from the job context, and searches all connected knowledge sources simultaneously: manuals, historical service records, prior case resolutions, and engineering bulletins. Within seconds, a response comes back directly in Teams with: * The most relevant diagnostic steps for that specific asset configuration * Citations showing exactly which sources the guidance came from * Links to supporting documentation for reference The engineer does not need to open any other system. The guidance is specific to the asset they are working on, not a generic procedure that requires interpretation. #### Follow-Up and Clarification The engineer follows the diagnostic steps and adds some additional context about what they're observing. The Knowledge Manager responds with further detail, continuing the conversation naturally and maintaining the context of the original question throughout. Each response can be rated with a thumbs up or thumbs down. This feedback is captured automatically and contributes to improving future responses for similar queries. #### Issue Resolved The engineer follows the recommended steps and resolves the issue on the first visit. The interaction has replaced what would previously have been 15 to 20 minutes of searching across multiple systems, a phone call to a senior technician, and a potential follow-up visit due to incomplete or generic guidance. #### Job Notes Generated When the job is closed, the Knowledge Manager can generate a set of job notes from the interaction, structured from the conversation, the guidance used, and the outcome. The engineer reviews and submits them. These notes contribute to the knowledge base, ensuring that the resolution is captured in a reusable form rather than buried in free-text. ### Scenario 2: Handling a Knowledge Gap This scenario covers what happens when a question arrives that the Knowledge Manager cannot answer with high confidence, for example, a query about a non-standard configuration for which documentation is incomplete or missing. #### Question Submitted An engineer submits a question about safety or configuration steps for a non-standard installation. The Knowledge Manager interprets the query and begins searching connected sources. #### Low Confidence Detected The system finds some relevant content but identifies that the available guidance has low confidence for this specific configuration, perhaps because documentation is partial, conflicting, or has not been updated to reflect recent equipment changes. Rather than returning the low-confidence guidance as if it were reliable, the Knowledge Manager does two things: 1. Requests additional context from the engineer, asking a targeted follow-up question to narrow the scenario before returning guidance. This ensures that if a reliable answer exists, it can be retrieved more precisely. 2. Flags the gap in the background, registering that this configuration has been queried and that the available knowledge is insufficient to answer it with confidence. #### Guidance Returned with Appropriate Signals If a high-confidence answer can be assembled after clarification, it is returned with citations and confidence indicators. If the gap is genuine and cannot be resolved from current sources, the Knowledge Manager communicates this clearly rather than returning unreliable guidance, and escalation paths are available for cases that require senior review. #### Gap Triggers Content Action The flagged gap is logged automatically. If the same configuration is queried repeatedly without a satisfactory answer, the Knowledge Manager surfaces this pattern for review, identifying it as a knowledge gap that needs to be addressed. It can generate a draft knowledge article based on the interaction context, which is routed through a human approval process before being published back into the knowledge base. Over time, this means the knowledge base evolves based on what engineers are actually encountering in the field, not just what was documented when the system was first set up. ## What These Scenarios Illustrate **Knowledge is delivered in the flow of work.** Engineers don't leave Teams to find answers. The entire interaction, question, guidance, follow-up, job notes, happens in the same interface they're already using. **Guidance is specific, not generic.** The Knowledge Manager uses asset ID, configuration, service history, and job context to return steps that apply to the exact situation. Generic documentation that might not apply is filtered out. **Citations make trust explicit.** Every response includes sources, so engineers can see where the guidance comes from and assess its authority. Confidence indicators make the reliability of each answer visible. **The system asks rather than guesses.** When context is missing or confidence is low, the Knowledge Manager asks a targeted follow-up question rather than returning uncertain guidance. This is particularly important in safety-sensitive environments where acting on incorrect information carries real risk. **Gaps are captured, not ignored.** When knowledge doesn't exist or is insufficient, the system flags it automatically rather than letting the gap persist. Over time this drives continuous improvement of the knowledge base from real field demand. **Resolved jobs feed future answers.** Knowledge generated through interactions, job notes, resolution steps, workarounds, is structured and routed for review rather than left in free-text notes. This turns individual resolutions into reusable organizational knowledge. # FAQ ## Capabilities & Functionality **Q: What does the Material Replenisher do?** A: The Material Replenisher is a Digital Worker that handles part requests from field technicians and maintenance teams. It interprets incoming requests from multiple channels, identifies and validates the correct part, checks inventory across connected sites, and either reserves the part, creates a material requisition, or routes the request to a planner depending on what the inventory check finds. The ERP and relevant stakeholders are updated automatically throughout the process. **Q: What types of requests can it interpret and act on?** A: The Digital Worker can interpret requests submitted via Microsoft Teams, email, work orders, IoT fault alerts, and customer calls or portals. It extracts part needs, urgency, and equipment context from the request, including from minimal or informal descriptions, and determines the appropriate action without requiring technicians to fill in forms or log into systems. **Q: How does it handle incomplete or vague part requests?** A: When a request does not contain enough information to identify a part with confidence, the Digital Worker asks a targeted clarifying question to gather what it needs. It does not reject the request or require the technician to start over. If multiple candidate parts match the description, it presents the options and asks the technician to confirm the correct one before proceeding. **Q: What happens when a part is in stock versus out of stock?** A: When a part is confirmed in stock at an accessible location, the Digital Worker reserves it and notifies the technician with the part details and location. When a part is not available locally, it conducts a multi-site inventory search across all connected locations. If the part cannot be found anywhere in inventory, the Digital Worker initiates a material purchase requisition (MPR) or material movement requisition (MMR) and guides the technician through the details needed to complete it. **Q: Can it identify and recommend similar parts?** A: Yes. The Digital Worker can suggest similar or equivalent parts based on minimal descriptions, historical usage data, and situational context. This reduces the risk of a request stalling because a technician does not have an exact part number. ## Control & Collaboration **Q: Does a human need to be involved in every request?** A: No. Requests with a clear part identification and available stock are handled automatically without planner involvement. The Digital Worker reserves the part, updates the ERP, and notifies the technician. Human input is required only when information is missing or ambiguous, when stock is unavailable and a requisition needs to be created, or when a request involves a judgment call that falls outside automated resolution. **Q: How does the Digital Worker involve planners when exceptions occur?** A: When a request cannot be resolved automatically, the Digital Worker routes it to a planner with the full context of the request, including the part details, equipment information, site, and the reason it requires review. Planners receive enough information to act quickly without having to investigate from scratch. **Q: How do technicians interact with the Digital Worker?** A: Technicians interact through natural conversation in Microsoft Teams. They can submit a request using plain language, respond to clarifying questions, confirm part selections, and provide requisition details, all within the same channel. No forms, no system logins, and no separate tools are required. **Q: What happens after a requisition is created?** A: Once a material requisition is created, the Digital Worker provides the technician with confirmation and next steps. It can also notify additional stakeholders if needed. The requisition details, decisions, and any approvals are recorded in full for audit purposes. ## Integration & Technology **Q: What systems does the Material Replenisher integrate with?** A: The Material Replenisher integrates with ERP and FSM systems for inventory checks, reservations, and work order updates. It connects to email platforms, collaboration tools including Microsoft Teams and Slack, and document platforms. This allows it to receive requests from existing channels and update records in the systems your teams already use. **Q: What channels can technicians use to submit requests?** A: Requests can be submitted via Microsoft Teams, email, work orders or work tasks in FSM and ERP systems, IoT fault alerts, and customer calls or portals. The Digital Worker monitors all of these simultaneously and processes requests in real time regardless of which channel they arrive through. **Q: How does it interact with the ERP?** A: The Digital Worker reads inventory data, validates parts against the catalog, creates or updates reservations, initiates MPRs or MMRs, and updates work order and work task records directly in the ERP and FSM. All updates are written back automatically without manual data entry. **Q: What is an MPR or MMR and how does the Digital Worker handle them?** A: An MPR (material purchase requisition) is created when a part needs to be sourced externally because it is not available in any connected inventory location. An MMR (material movement requisition) is created when stock needs to be transferred between sites. The Digital Worker initiates the appropriate document type based on the inventory check outcome, collects the required details from the technician, and creates the record in the ERP automatically. ## Security & Reliability **Q: How does the Digital Worker prevent duplicate orders or incorrect part selection?** A: Parts de-duplication is a core agentic skill. The Digital Worker checks whether a request duplicates an existing reservation or order before creating anything new. When multiple parts match a description, it presents all candidates for technician confirmation rather than selecting one automatically, preventing incorrect reservations from entering the system. **Q: How is traceability maintained across requests?** A: Every request, part validation, inventory check, reservation, MPR or MMR, and approval is recorded with full traceability. This means planners and auditors can review the complete history of any request, including what information was provided, what decisions were made, and what actions were taken. **Q: How are exceptions handled consistently across sites?** A: Exception routing follows the same logic regardless of which site a request originates from. When a request cannot be resolved automatically, it is escalated to a planner with standardized context. This means all sites operate on the same intake process with no local variations in how exceptions are handled. ## Customization & Performance **Q: Can it operate across multiple sites?** A: Yes. The Material Replenisher checks inventory across all connected sites simultaneously and applies the same intake and validation logic at every location. Organizations managing replenishment across many sites benefit from consistent processes and consolidated visibility without additional configuration per site. **Q: What agentic skills does the Material Replenisher use?** A: The Material Replenisher is built on five agentic skills: Request Interpretation, Part Identification & Validation, Multi-Site Inventory Lookup, Fulfilment Decision Logic, Parts De-duplication, and Requisition Creation. Together these skills cover the full path from an incoming technician request through to ERP update or requisition creation. **Q: Can it be configured to reflect our specific parts catalog and business rules?** A: Yes. The Digital Worker validates parts against your catalog in the connected ERP and applies fulfillment logic based on your inventory structure and site configuration. This means part identification and requisition creation reflect your organization's actual data rather than generic rules. # Case studies See how our customers are using this Digital Worker to transform their operations. {% embed url="" %} # Supplier Invoicing Manager Invoice processing is detail-intensive work — and the time your team spends chasing approvals, matching POs, and resolving discrepancies is time better spent elsewhere. This knowledge base covers everything you need to get the most out of the Supplier Invoicing Manager, from setup and configuration to day-to-day use. ### Jump right in
Cover image
Watch & learnSee this Digital Worker in action./files/IniyzVQmRw3Net0b8IkX/pages/WNsSf5SPFS2a7kCcNpP7
Core conceptsEssential context for understanding how it all works./files/fZNxfxVnHxx6iNh5M9k9/pages/JLaDC4X9KQk3varKOFv1
FAQQuick answers to common questions./files/RBUXYX1hxsMuVshLZ5Ar/pages/bBsJowtbBVt8z4TqclKP
How-to guidesPractical, step-by-step guidance across a range of tasks/files/uC1RptEfw6kewkOX3sMG/pages/2Qs8NANzzJTvoXvABd7B
# Watch & learn Watch the video below to see this Digital Worker in action {% embed url="" %} # Core concepts Explore core concepts to build a solid understanding of the Digital Worker. These articles will give you the context needed to get the most out of your Digital Worker.
Cover image
FundamentalsLearn the full scope of what's possible/files/PKWzoCypd9fEWl1gt5o8/pages/KTAnbOQm7KVnUPoxc5A8
Architecture & capabilitiesDiscover the components and integrations/files/MgisEby9ZvW43Wybsymq/pages/bkctjjz2tIkdPi9ywy0A
Real-world scenariosPractical examples of the platform in action across common business use cases./files/I8b6oW47o0Zv7grfRIxS/pages/DwZ80OYCyYco8xhyEMKA
# Fundamentals ## Overview The Supplier Invoicing Manager is an AI-powered Digital Worker designed to free Accounts Payable teams from the manual, repetitive work of processing supplier invoices so they can focus on exceptions and strategic priorities. It automates invoice intake, data extraction, three-way matching against purchase orders and receipts, and exception routing, ensuring invoices are validated accurately and approved quickly, with full audit traceability throughout. By transforming manual, fragmented invoice processing into a structured, automated workflow, the Supplier Invoicing Manager gives AP teams time back for the decisions that require human judgment, while reducing errors, preventing duplicate payments, and improving compliance. ## Key Benefits **Automated invoice intake:** Supplier invoices arriving via email are ingested and parsed automatically, no manual opening of attachments, no re-entering of data, no routing decisions required for clean invoices. **Three-way matching:** Every invoice is validated against the corresponding purchase order and goods receipt note before approval, catching discrepancies in price, quantity, and totals before they reach the ERP. **Intelligent variance interpretation:** The system interprets service charges, taxes, freight, and price or quantity variances, even when part numbers are missing or invoice formats differ across suppliers, rather than simply flagging a mismatch and stopping. **Auto-approval for valid invoices:** Invoices that pass all matching checks within configured tolerance thresholds are approved automatically and the ERP is updated instantly, with the invoice copy attached. No human touchpoint required. **Exception routing with full context:** When a discrepancy is identified, the invoice is routed to the right approver with a clear explanation of the issue, so resolution takes seconds rather than requiring manual investigation. **Audit-ready traceability:** Every decision, action, and communication is recorded with full traceability, ensuring discrepancies and resolutions are documented consistently for compliance and audit purposes. ### The Challenge It Solves Supplier invoicing is often manual, fragmented, and exception-heavy. AP teams must extract invoice data from emails and documents, perform two-way or three-way matching across systems, interpret variances, and resolve discrepancies through back-and-forth communication with suppliers. When invoice lines don't match what was ordered or received, or when service charges and variances are unclear, approvals slow down and exceptions pile up. The result is delayed payment processing, avoidable rework, duplicate invoicing risk, and increased exposure to late payment penalties or missed early payment discounts. Teams spend significant time on tasks that are rules-based and repetitive rather than on the financial analysis and supplier relationship work that requires their expertise. ## Before and After ### Before Before the Supplier Invoicing Manager, processing a supplier invoice requires manual effort at every stage: {% stepper %} {% step %} Process inbox and review invoice PDFs and attachments {% endstep %} {% step %} Manually match the invoice against the purchase order {% endstep %} {% step %} Validate against the goods receipt note {% endstep %} {% step %} Validate pricing and tolerance thresholds {% endstep %} {% step %} Update the ERP {% endstep %} {% step %} Manage exceptions through email back-and-forth {% endstep %} {% step %} Track and follow up on outstanding resolutions {% endstep %} {% endstepper %} This process is slow, inconsistent across team members, and difficult to audit when discrepancies occur. ### After With the Supplier Invoicing Manager, the same cycle runs in three automated steps:
{% stepper %} {% step %} Monitor the AP inbox continuously and extract invoice data {% endstep %} {% step %} Match the invoice against the PO and receipt using three-way matching, validate pricing and tolerances {% endstep %} {% step %} Match the invoice against the PO and receipt using three-way matching, validate pricing and tolerances {% endstep %} {% endstepper %} Clean invoices are processed without any human involvement. Exceptions reach the right person immediately, with everything they need to resolve the issue quickly. ## How It Works The Supplier Invoicing Manager executes a structured automated workflow that handles the complete invoice processing lifecycle. **Invoice Intake and Parsing:** The Digital Worker monitors the AP inbox continuously, capturing supplier invoices as they arrive via email. It ingests the invoice, extracts header and line-level details, invoice number, supplier, amounts, dates, PO references, and line items, and converts unstructured document content into structured financial fields ready for validation. **Three-Way Matching:** The system validates the invoice by comparing it line by line against the corresponding purchase order and goods receipt note. It checks for accuracy in quantity, price, and totals, and identifies any mismatches that fall outside configured tolerance thresholds. **Charge and Variance Interpretation:** When variances are detected, the system interprets them in context, classifying service charges, freight, taxes, and line-level discrepancies, and determining what can be resolved automatically versus what requires human review. This goes beyond simple flag-and-stop logic to provide meaningful context about what the variance represents. **Auto-Approval and ERP Update:** Invoices that pass all checks within tolerance thresholds are approved automatically. The ERP is updated instantly with the invoice status, and the invoice copy is attached to the record. No manual step is required. **Exception Routing:** Invoices with unresolvable discrepancies are routed to the appropriate AP team member with a clear explanation of the issue and the decision options available. Once the reviewer provides their decision, the Digital Worker handles the follow-up, including contacting the supplier directly to flag the discrepancy and request a correction. # Architecture & capabilities ## Overview The Supplier Invoicing Manager is built on four integrated agentic skills coordinated by a Supervisor Agent (AP & Invoice Processing Manager). This article covers the core skills, the full invoice processing workflow, integration points, and how the system determines when to auto-approve versus when to route for human review. ## Agentic Skills The Supplier Invoicing Manager operates through four core skills, each handling a distinct stage of the invoice processing lifecycle. #### Invoice Intake & Parsing Automatically receives supplier invoices from inbound channels and extracts header and line-level details such as invoice number, amounts, dates, PO references, and line items. This skill handles variability in invoice formats across suppliers, PDFs, scanned images, email body content, and converts unstructured input into structured financial fields that can be validated against ERP records. Invoices arriving from multiple channels are captured and prepared for processing without any manual data entry. #### 3-Way Match Validates invoices by matching them against the corresponding purchase order and goods receipt note to ensure accuracy in quantity, price, and totals. The match is performed line by line, using configurable tolerance logic to determine whether a variance falls within acceptable bounds or requires escalation. Where all three records align within tolerance, the invoice is cleared for auto-approval. Where they do not, the mismatch is passed to the Charge & Variance Interpretation skill before any escalation decision is made. #### Charge & Variance Interpretation Identifies service charges, taxes, freight, and any price or quantity variances, even when part numbers are missing or invoice formats differ from what the system expects. Rather than treating all variances as equivalent exceptions, this skill classifies each one: determining what the charge represents, whether it has a legitimate basis, and whether it can be resolved automatically or requires human review. This context-aware interpretation reduces the volume of exceptions that reach AP teams by resolving variances that are explainable within the invoice's context. #### Exception Handling Routes mismatches or unclear cases to the right approver with a clear explanation of the issue, enabling fast resolution without manual investigation. When an exception is escalated, the reviewer receives the invoice details, the specific discrepancy, and the decision options available. Once a decision is made, the Digital Worker executes the follow-up automatically, including contacting the supplier directly to flag the issue and request a correction if needed. All exceptions are routed consistently using defined matching rules and tolerance thresholds, ensuring no discrepancy is handled ad hoc or inconsistently. ### Invoice Processing Workflow The following table outlines the full invoice processing workflow from initial intake through to ERP update and supplier communication. | **Stage** | **What the System Does** | **Action / Output** | **Channel** | | ---------------------------------------- | ----------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------- | ----------------- | | **Invoice Intake & Parsing** | Parses invoices from email and PDF; extracts supplier, PO reference, amounts, and line items | Structures invoice data, validates completeness, triggers processing workflow | Email / OCR / IFS | | **3-Way Matching** | Compares invoice against PO and goods receipt; identifies mismatches in price, quantity, and delivery | Validates invoice accuracy, auto-matches where possible, flags discrepancies | IFS | | **Exception Detection & Classification** | Identifies mismatches and tolerance breaches; classifies severity and type of issue | Triggers exception workflows, prioritizes by impact, prevents incorrect posting | IFS / Teams | | **Invoice Posting & ERP Update** | Validates final data against rules; ensures readiness for posting | Posts invoice to IFS Cloud, maintains audit trail, ensures financial accuracy | IFS | | **Supplier Communication & Resolution** | Detects need for clarification; identifies missing or incorrect invoice details | Drafts supplier emails, requests corrections, tracks responses | Email | | **Human-in-the-Loop Decision Support** | Identifies ambiguous scenarios; provides context for decisions | Surfaces decision options to reviewer, captures approvals or rejections, resumes automated flow after decision | Teams | ## Integration Points **Email Systems:** Monitors Microsoft Outlook and Gmail inboxes continuously, capturing supplier invoices with PDF and document attachments in real time. **ERP Platforms:** Integrates directly with IFS Cloud to read purchase order and goods receipt records, validate invoice data, post approved invoices, and update invoice status automatically. **Collaboration Platforms:** Uses Microsoft Teams and Slack for exception notifications, approval workflows, and human-in-the-loop decision support. **Documents:** Integrates with IFS Loops and Google Drive for document storage and invoice record management. **Enterprise Applications:** Connects with IFS and other enterprise platforms for cross-system data access and financial record management. ## Human-in-the-Loop Design The Supplier Invoicing Manager is designed so that clean invoices are processed without any human touchpoint, and human judgment is applied only where genuine decisions are required. The system proceeds independently when: * Invoice details match the purchase order and goods receipt within configured tolerance thresholds * Variances are interpretable as legitimate charges (freight, tax, service fees) within tolerance * The invoice has not been processed before (idempotency check passes) * All required fields are present and valid Human review is required when: * Price, quantity, or total mismatches exceed configured tolerance thresholds * Variances cannot be classified or resolved automatically * Invoice references are missing, ambiguous, or cannot be matched to a PO * Duplicate invoice risk is detected * The discrepancy requires a business decision before the supplier can be contacted When an exception is routed for review, the system sends a notification with the invoice details, the specific issue, and the rationale, so the reviewer has everything needed to make a decision without further investigation. Once approved or rejected, the Digital Worker immediately executes the appropriate next step, including any required supplier communication. ## Audit Traceability Every invoice processed by the Supplier Invoicing Manager generates a complete record of the decisions, validations, and actions taken. This includes the matching outcome for each line item, any variances detected and how they were classified, whether the invoice was auto-approved or escalated, the decision made by the human reviewer if applicable, and any supplier communications sent. This structure ensures that AP operations are fully auditable and that discrepancies are resolved with documented rationale rather than informal resolution. # Real-world scenarios ## Overview This article walks through two practical scenarios that illustrate how the Supplier Invoicing Manager handles incoming supplier invoices. The first covers a clean invoice that passes all matching checks and is approved without any human involvement. The second covers an invoice where a discrepancy is detected, a human reviewer is brought in to make a decision, and the Digital Worker handles the supplier follow-up automatically. ### Scenario 1: Clean Invoice, Auto-Approval This scenario covers the most common type of interaction: a supplier submits an invoice that matches the corresponding purchase order and goods receipt note within configured tolerance thresholds. **What triggers it:** A supplier invoice arrives via email to the AP team's monitored inbox, with the invoice attached as a PDF. **What the Digital Worker does:** 1. The Invoice Intake & Parsing skill captures the email, extracts the invoice document, and converts the header and line-level details, invoice number, supplier, amounts, dates, PO reference, and line items, into structured financial fields. 2. The 3-Way Match skill retrieves the corresponding purchase order and goods receipt note from IFS Cloud and compares them against the invoice line by line, checking quantities, prices, and totals. 3. No discrepancies are found. All values fall within configured tolerance thresholds. 4. The invoice is auto-approved. The ERP is updated instantly with the invoice status, and the invoice copy is attached to the record. **Result:** The invoice is fully processed, from email receipt to ERP update, without any manual intervention. The AP team receives visibility into what was processed without needing to touch the invoice themselves. What would have previously required opening the attachment, locating the PO, performing the match manually, and updating the system is handled entirely in the background. ### Scenario 2: Invoice with a Discrepancy, Exception Routing and Supplier Follow-Up This scenario covers an invoice where the amount on the supplier's invoice does not match the corresponding purchase order. The Digital Worker detects the mismatch, routes it to a human reviewer with full context, and handles the supplier communication once a decision is made. #### Step 1: Invoice Received and Parsed A supplier invoice arrives via email. The Digital Worker picks it up immediately, extracts the invoice data, and begins the three-way matching process, comparing the invoice against the purchase order and goods receipt records in IFS Cloud. #### Step 2: Discrepancy Detected During matching, the system identifies that the invoice amount does not match the purchase order. Rather than letting the discrepancy move forward unchecked, or rejecting the invoice outright, the Digital Worker flags the specific mismatch and classifies it for review. The Charge & Variance Interpretation skill assesses the variance in context. If the discrepancy is explainable as a legitimate charge within tolerance (such as a freight fee or tax variance), it may be resolved automatically. In this case, the numbers don't reconcile, and the issue is passed to Exception Handling for human review. #### Step 3: Exception Routed to Reviewer The AP team receives a notification, via Microsoft Teams or the configured channel, with the invoice details and a clear explanation of what doesn't match. The reviewer can see the original PO amount, the invoiced amount, and the specific line items involved. Everything needed to make a decision is surfaced in the notification; no additional investigation is required. #### Step 4: Reviewer Provides Decision The reviewer assesses the situation and provides their decision. This takes seconds rather than the time it would have taken to locate the relevant records, investigate the discrepancy, and compose an email to the supplier manually. #### Step 5: Supplier Communication Handled Automatically Once the reviewer's decision is captured, the Digital Worker takes over. It emails the supplier directly to flag the discrepancy and request a correction, no manual inbox monitoring, no composing of emails from scratch. The communication is generated based on the invoice context and the reviewer's decision, sent immediately, and tracked. **Result:** The discrepancy was caught before it reached the ERP. The reviewer made a fast, informed decision with full context in front of them. The supplier was contacted automatically with a clear, professional message. The entire chain of events, detection, escalation, decision, and supplier communication, is recorded for audit purposes. #### What These Scenarios Illustrate **Clean invoices require no human involvement.** When an invoice matches the PO and receipt within tolerance, the Digital Worker handles extraction, matching, approval, and ERP update end to end. AP teams are informed without needing to act. **Discrepancies are caught before they reach the ERP.** The three-way match happens during the validation stage, before any data is posted. This prevents incorrect financial records from being created and avoids the rework that would follow an incorrect posting. **Variance interpretation reduces unnecessary escalations.** Not every mismatch becomes an exception. The system interprets charges in context, classifying freight, taxes, and service fees, and resolves what it can within configured tolerance before routing anything to a human. **Reviewers get decisions, not investigations.** When an exception does reach a human, it arrives with the relevant details, the specific discrepancy, and the decision options already laid out. Reviewers focus on the judgment call, not on finding the information needed to make it. **Supplier follow-up is handled automatically.** Once a decision is made, the Digital Worker manages the supplier communication, drafting and sending the message based on the invoice context and the decision provided. Teams stay focused on decisions, not admin. **Every action is auditable.** The full chain, intake, matching, variance classification, escalation, decision, and communication, is recorded consistently for every invoice processed. There is no gap in the audit trail regardless of whether the invoice was auto-approved or required human review. # FAQ ## Capabilities & Functionality **Q: What does the Supplier Invoicing Manager do?** A: The Supplier Invoicing Manager is a Digital Worker that automates supplier invoice intake, validation, and exception routing for accounts payable teams. It ingests invoices received by email, parses and extracts the invoice data, validates each invoice against the corresponding purchase order and goods receipt, and either auto-approves the invoice or routes it for human review based on the outcome of that validation. ERP records are updated automatically once an invoice is approved. **Q: What is three-way matching and how does the Digital Worker perform it?** A: Three-way matching is the process of comparing an invoice against both the originating purchase order and the goods receipt to confirm that what was invoiced matches what was ordered and received. The Supplier Invoicing Manager executes this comparison automatically for every invoice it processes. It applies configurable tolerance logic to determine whether any differences fall within acceptable thresholds or require escalation. **Q: How does the Digital Worker handle invoices that do not match the PO or receipt?** A: When the Digital Worker detects a discrepancy that falls outside the defined tolerance thresholds, it flags the mismatch and routes the invoice to an AP team member for review. The reviewer receives the relevant details and context needed to assess the situation. Once they provide a decision, the Digital Worker takes follow-up action, including emailing the supplier to request a correction if needed. **Q: What happens after a valid invoice is approved?** A: Valid invoices are auto-approved without requiring manual touchpoints. The ERP is updated immediately to reflect the approval, and a copy of the invoice is attached to the record. The entire process runs without human intervention for invoices that match within tolerance. **Q: Can the Digital Worker handle service charges and line-item variances?** A: Yes. Charge and variance interpretation is one of the Digital Worker's core capabilities. It classifies variances across line items and determines whether each can be resolved within configured thresholds or requires escalation for review. This prevents unclear charges from stalling the approval process or slipping through unchecked. ### Control & Collaboration **Q: Does a human review every invoice?** A: No. Invoices that match within configured tolerance thresholds are auto-approved without any manual review. Human involvement is reserved for genuine exceptions, specifically invoices where discrepancies, unclear charges, or variances cannot be resolved automatically. This keeps your AP team focused on decisions that require judgment rather than routine processing. **Q: How does the Digital Worker involve AP team members when an exception occurs?** A: When an invoice requires human review, the Digital Worker routes it to the appropriate team member with the relevant details surfaced so they can act quickly and with full context. The reviewer assesses the situation and provides their feedback, after which the Digital Worker handles any required follow-up. **Q: What information does the reviewer receive when an invoice is escalated?** A: Escalated invoices are accompanied by the details needed to understand the discrepancy, including the invoice data, the relevant purchase order and receipt records, and the nature of the mismatch or variance. Reviewers can make informed decisions without having to manually locate supporting documents. **Q: How does the Digital Worker communicate with suppliers?** A: After a reviewer assesses a discrepancy and provides feedback, the Digital Worker sends an email directly to the supplier to flag the issue and request a correction. This removes the need for manual inbox monitoring or back-and-forth coordination by the AP team. ## Integration & Technology **Q: What systems does the Supplier Invoicing Manager integrate with?** A: The Supplier Invoicing Manager integrates with email platforms, enterprise applications including ERP and CRM systems, collaboration tools, and document platforms. This allows it to receive invoices from existing inbound channels and update records in the systems your team already uses. **Q: How does the Digital Worker receive invoices?** A: Invoices are captured from inbound email channels. The Digital Worker monitors these channels, ingests invoices as they arrive, and prepares them for processing automatically. No manual upload or intervention is required to initiate the process. **Q: How does it update the ERP after processing?** A: Once an invoice is approved, the Supplier Invoicing Manager updates the invoice status in the ERP directly. This keeps invoice progress and outcomes visible in your system of record without requiring manual data entry by the AP team. **Q: What collaboration and document platforms are supported?** A: The Digital Worker supports integration with collaboration platforms including Slack and Microsoft Teams, and document platforms including IFS and Google Drive. Refer to your implementation documentation for the full list of supported connectors available in your environment. ## Security & Reliability **Q: How is audit traceability maintained?** A: Every decision the Digital Worker makes is traceable. Discrepancies, the context in which they were detected, and how they were resolved are all recorded with full traceability. This means AP teams and auditors can review the rationale behind any approval or escalation after the fact. **Q: What prevents duplicate payments or overpayment?** A: The Digital Worker enforces consistent matching rules and applies idempotency controls to prevent invoices from being processed more than once. Combined with three-way matching against PO and receipt records, this significantly reduces the risk of duplicate payments and overpayment. **Q: How are exceptions handled consistently?** A: Exception routing follows defined tolerance and matching rules configured for your environment. Every invoice is evaluated against the same criteria, regardless of volume or who is processing the queue. Consistent rules also make it easier to audit exception decisions and identify patterns over time. ## Customization & Performance **Q: Can matching rules and tolerance thresholds be configured?** A: Yes. The Supplier Invoicing Manager uses configurable tolerance logic for three-way matching and auto-approval decisions. This allows organizations to define the thresholds that reflect their own procurement and finance policies, rather than applying fixed rules that may not suit their business. **Q: Can the Digital Worker be adjusted to route specific exception types differently?** A: Exception routing is governed by the rules configured in your environment. Invoices that fall outside tolerance thresholds are escalated with decision rationale, allowing your team to define how different types of discrepancies should be handled and by whom. **Q: What agentic skills does the Supplier Invoicing Manager use?** A: The Supplier Invoicing Manager is built on four agentic skills: Invoice Intake & Parsing, 3-Way Match, Charge & Variance Interpretation, and Exception Handling. Each skill handles a distinct part of the invoice processing lifecycle, and together they cover the full path from invoice receipt through to ERP update or supplier follow-up.
# Case studies See how our customers are using this Digital Worker to transform their operations. {% embed url="" %} # Supplier Order Manager Supplier order coordination is complex work — and most of it shouldn't require human effort. This knowledge base covers everything you need to get the most out of the Supplier Order Manager, from setup and configuration to day-to-day use. ### Jump right in
Cover image
Watch & learnSee this Digital Worker in action./files/IniyzVQmRw3Net0b8IkX/pages/TUwqRA5UgiDOyxHHes4z
Core conceptsEssential context for understanding how it all works./files/fZNxfxVnHxx6iNh5M9k9/pages/iPcVHiQhOOZixi6EMAw5
FAQQuick answers to common questions./files/RBUXYX1hxsMuVshLZ5Ar/pages/eB953vLfEaQvm6HQh8OX
Case studiesLearn from real world examples of success./files/c1GhIydX1yr8ZcJL60g6/pages/S7aS4Y33DMDXYc4GMgKN
How-to guidesPractical, step-by-step guidance across a range of tasks/files/uC1RptEfw6kewkOX3sMG/pages/2Qs8NANzzJTvoXvABd7B
# Watch & learn Watch the video below to see this Digital Worker in action {% embed url="" %} # Core concepts Explore core concepts to build a solid understanding of the Digital Worker. These articles will give you the context needed to get the most out of your Digital Worker.
Cover image
FundamentalsLearn the full scope of what's possible/files/PKWzoCypd9fEWl1gt5o8/pages/fpa07XiAgokkz23CKJjU
Architecture & capabilitiesDiscover the components and integrations/files/MgisEby9ZvW43Wybsymq/pages/EWGX9qaYvDTN06jJ9avy
Real-world scenariosPractical examples of the platform in action across common business use cases./files/I8b6oW47o0Zv7grfRIxS/pages/jE0Uvbw43E6DgdXFux5V
# Fundamentals ## Overview The Supplier Order Manager is an AI-powered Digital Worker designed to free procurement teams from the administrative work of managing supplier confirmations so they can focus on strategic sourcing, category management, and the relationship building that drives competitive advantage. This intelligent solution handles the complex workflow of confirming purchase orders from suppliers, detecting discrepancies, and updating system records with minimal manual intervention. By converting reactive, manual supplier management into a proactive, exception-driven process, the Supplier Order Manager gives procurement professionals time back for the work that requires human judgment, negotiation skills, and strategic thinking. ## Key Benefits **Automated confirmation handling:** Supplier confirmations are parsed, validated against purchase orders, and processed without manual review for clean matches, eliminating the need to open every email and compare it against the system manually. **Proactive exception detection:** Discrepancies in price, quantity, or delivery dates are caught automatically and surfaced to the right person with full context, rather than being discovered late or missed entirely. **Consistent supplier communication:** All outbound communications, confirmations, reminders, clarifications, are drafted and sent through a standardized, professional process rather than relying on individual team members to manage their own inboxes. **Real-time visibility:** Teams have an up-to-date view of confirmation status across all open purchase orders without manually chasing responses or checking system records. **Effortless scalability:** The automated process handles growing confirmation volume without additional manual effort, eliminating bottlenecks during periods of high order activity. ## The Challenge It Solves Managing supplier orders is highly manual, fragmented, and reactive. Procurement teams spend significant time tracking confirmations, validating changes to price, quantity, and delivery dates, and following up with suppliers across emails and ERP systems. This leads to delays, missed discrepancies, and limited visibility into order status, ultimately impacting supply continuity and operational efficiency. In practice, teams are doing tasks like opening supplier emails, reviewing attached PDFs, comparing details against purchase orders in IFS Cloud, identifying any differences, sending clarification or follow-up emails, and then manually updating the ERP once changes are agreed. When suppliers propose modified terms, the process involves even more back-and-forth before anything gets resolved. Multiply this across dozens or hundreds of confirmations per day and the administrative burden becomes a significant drain on procurement capacity. ## Before and After ### Before Before the Supplier Order Manager, processing a supplier confirmation requires manual effort at every step: {% stepper %} {% step %} Process inbox and review purchase order confirmation emails and attachments {% endstep %} {% step %} Compare confirmation against the purchase order in IFS Cloud {% endstep %} {% step %} Identify any discrepancies {% endstep %} {% step %} Send follow-ups or clarifications to the supplier {% endstep %} {% step %} Track supplier responses {% endstep %} {% step %} Update the ERP with any approved changes {% endstep %} {% step %} Send confirmation back to the supplier {% endstep %} {% endstepper %} This manual process leads to delayed confirmations, frequent back-and-forth emails, poor visibility into open order status, and error-prone ERP updates. ### After With the Supplier Order Manager, the same cycle is handled in three automated steps: {% stepper %} {% step %} Monitor inbox continuously and extract purchase order-related data {% endstep %} {% step %} Match confirmations with ERP records, detect and flag any changes to a human for review {% endstep %} {% step %} Handle supplier communication and update the ERP with confirmed changes {% endstep %} {% endstepper %} Exact-match confirmations require no human involvement at all. Exceptions are routed immediately with all the context needed to make a fast, informed decision. ## How It Works The Supplier Order Manager executes a six-step automated workflow that covers the complete purchase order confirmation cycle. **Information Extraction:** The Digital Worker extracts all relevant information from the supplier's purchase order confirmation, whether it arrives as a PDF, image, or through EDI. An AI-powered extraction tool converts unstructured document data into structured format with key-value pairs. **Details Retrieval:** A specialized Purchase Order Details sub-agent retrieves the original purchase order from IFS Cloud, including site information, line items for each part number, quantities, prices, delivery dates, and all relevant metadata. **Intelligent Comparison:** The system performs a line-by-line comparison between the supplier's confirmation and the original purchase order. It checks for duplicate requests, uses LLM-based comparison to identify any differences, and determines whether to confirm as-is or flag discrepancies for review. **Exception Routing:** When discrepancies are detected, the system triggers a human approval workflow. A detailed notification is sent via Microsoft Teams or the configured channel, outlining exactly what changed and waiting for explicit approval before proceeding. **System Update:** Once approved, the Digital Worker automatically updates all relevant purchase order line items in IFS Cloud to reflect the confirmed changes. **Supplier Communication:** The system drafts and sends confirmation back to the supplier, maintaining consistent, professional communication throughout. ## Purchase Order States When a purchase order is created in IFS Cloud, it begins in a planned state. Once sent to the supplier, it moves to a released state. The Supplier Order Manager monitors continuously for incoming supplier responses and processes them as they arrive, handling both types of supplier response: * **Full acknowledgment:** The supplier agrees with all terms and confirms as submitted * **Modified acknowledgment:** The supplier proposes changes such as reduced quantities, adjusted prices, or revised delivery dates
# Architecture & capabilities ## Overview The Supplier Order Manager is built on a hierarchical AI agent architecture with a set of integrated agentic skills coordinated by a Supervisor Agent. This article covers the core skills, the agent architecture, the full supplier order management workflow, integration points, and how the system determines when to proceed independently versus when to involve a human. ## Agentic Skills The Supplier Order Manager operates through five core skills, each handling a distinct part of the supplier order management process. #### Email Parsing Reads and understands supplier emails and attachments to extract confirmations, changes, and intent without manual review. This skill processes incoming messages regardless of format, plain text, PDF attachments, images, and converts unstructured supplier communication into structured data that can be compared against purchase order records in IFS Cloud. #### Supplier Coordination Keeps suppliers aligned by tracking responses, follow-ups, and next steps across ongoing purchase orders. This skill maintains visibility into which confirmations are outstanding, which have been received, and which require further action, ensuring nothing falls through the cracks across a high volume of open orders. #### Supplier Communication Manages structured, consistent communication with suppliers to prevent delays, ambiguity, and missed information. This skill ensures all supplier-facing messages follow a standardized approach, reducing the variability that comes from individual team members managing their own inboxes and writing their own emails. #### Email Composition Drafts clear, professional emails for confirmations, reminders, and clarifications based on the current purchase order context. Rather than starting from a blank page, this skill generates contextually appropriate messages ready for review and sending, whether acknowledging a clean confirmation, requesting clarification on a discrepancy, or following up on an overdue response. #### Supplier Negotiation Supports back-and-forth discussions on dates, quantities, or commercial terms by capturing proposed changes, highlighting trade-offs, and escalating when approval is needed. This skill handles the structured aspects of negotiation, documenting what's been proposed, what the impact is, and what decision is required, so that human reviewers can focus on the judgment call rather than the administrative tracking. ## Agent Architecture The Supplier Order Manager uses a hierarchical agent structure designed for reliability, scalability, and transparency. **Main Agent (Supervisor):** The Supplier Order Manager serves as the orchestration layer, overseeing the entire process from email receipt through system update and supplier confirmation. It directs the appropriate skills based on what each incoming communication requires. **Instruction Handlers:** Specialized language models handle specific tasks such as data extraction from documents and intelligent comparison of parts and pricing. These handlers provide the AI reasoning capability that makes the system adaptable to various document formats and supplier communication styles. **Sub-Agents:** Dedicated AI skills focus on purchase order details retrieval and confirmation management. These sub-agents have deep expertise in their specific domains, ensuring accurate execution of critical workflow steps. ## Tools and Integration Layer The Supplier Order Manager uses three core configurable tools: * **Extract Order Confirmation Details:** an instruction handler that extracts structured data from supplier purchase order confirmations * **Compare Parts Tool:** an LLM-based tool that compares extracted parts with system records, identifying mismatches in quantities, prices, and attributes * **Teams Integration:** configurable channels for agent communication and human-in-the-loop approval workflows ### Supplier Order Management Workflow The following table outlines the full workflow from initial confirmation receipt through to ERP update and supplier communication. | **Stage** | **What the System Does** | **Action / Output** | **Channel** | | ---------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------- | ----------------- | | **Supplier Confirmation Parsing & Intent Detection** | Parses supplier confirmations from PDF or email; extracts PO, line items, quantities, dates, and pricing; detects whether the response is a full confirmation, partial confirmation, or rejection | Structures confirmation data against PO, classifies confirmation type, triggers next workflow step | Email / OCR / IFS | | **PO Matching & Change Detection** | Compares confirmation against PO line-by-line; detects changes in price, quantity, and delivery date; identifies missing or extra lines | Highlights discrepancies automatically, classifies changes as acceptable or exception, routes to appropriate workflow | IFS / Teams | | **Exception Detection & Prioritization** | Identifies critical vs. non-critical mismatches; applies tolerance rules for price, date, and quantity; evaluates fulfilllment impact | Triggers exception workflows, notifies buyers for decision, prevents incorrect confirmations from posting | Teams / Email | | **Supplier Communication & Follow-ups** | Detects missing confirmations or delayed responses; identifies need for clarification or negotiation | Sends automated follow-ups and reminders, drafts clarification emails, initiates negotiation flows | Email | | **Confirmation Posting & ERP Updates** | Validates confirmed data against business rules; ensures readiness for ERP update | Posts confirmed updates to IFS Cloud, maintains audit trail of changes, ensures data consistency | IFS | | **Human-in-the-Loop Decision Support** | Identifies ambiguous or conflicting scenarios; provides context for decision-making | Surfaces clear options to reviewer, highlights risks and impact, captures decision and resumes automated flow | Teams / IFS | ## Integration Points **Email Systems:** Monitors Microsoft Outlook and Gmail inboxes continuously, processing incoming supplier communications in real time. **ERP/SCM Platforms:** Integrates directly with IFS Cloud and other enterprise resource planning systems to read purchase order details and write updates automatically. **Collaboration Platforms:** Uses Microsoft Teams, Slack, OneNote, and Google Drive for notifications, approvals, and documentation. **CRM Systems:** Connects with Salesforce, HubSpot, and Qoncierge to maintain supplier relationship context and history. **Document Processing:** Handles PDFs, images (JPG, PNG), and text-based confirmations with high extraction accuracy. ## Human-in-the-Loop Design The Supplier Order Manager is designed so that routine confirmation processing happens automatically and human judgment is applied to decisions that genuinely require it. The system proceeds independently when: * Supplier confirmations match purchase orders exactly across all line items * Communications follow established patterns and require no exception handling * Transactions are low-risk and fall within approved parameters Human approval is required when: * Any discrepancy exists in dates, quantities, or pricing * High-value or high-impact purchase orders are involved * Cases require business judgment that falls outside defined rules * Outbound supplier communications are ready to be sent When an exception is detected, the system sends a detailed notification via Microsoft Teams with complete context, presents a clear comparison showing exactly what changed and why it matters, waits for explicit approval before taking any action, and executes the approved changes immediately once confirmation is received. ## Testing and Validation The platform includes an agent playground for testing and development. Teams can simulate incoming emails with PDF attachments, test responses to various scenarios, review detailed debug logs, and validate extraction accuracy and comparison logic before deploying to production.
# Real-world scenarios ## Overview This article walks through two practical scenarios that illustrate how the Supplier Order Manager handles incoming supplier confirmations. The first covers a clean confirmation where all details match the purchase order exactly. The second covers a confirmation where the supplier has proposed a change to the delivery date, requiring human review before the system can proceed. ### Scenario 1: Exact Match Confirmation This scenario covers the most common type of interaction, a supplier confirms they can fulfilll all line items at the agreed quantities, prices, and delivery date. Everything aligns with the purchase order on record. **What triggers it:** A supplier sends a purchase order confirmation via email to the order manager's monitored inbox. **What the Digital Worker does:** 1. The Email Parsing skill detects the incoming message, reads the email and any attached documents, and extracts all relevant confirmation details. 2. The system retrieves the original purchase order from IFS Cloud, line items, quantities, prices, delivery dates, and all associated metadata. 3. A line-by-line comparison is performed between the supplier's confirmation and the purchase order. No discrepancies are found. 4. The order is confirmed automatically in IFS Cloud without any manual intervention. 5. The procurement team receives a notification via Microsoft Teams that the purchase order has been confirmed. **Result:** The entire confirmation process, email receipt, data extraction, PO comparison, ERP update, and team notification, happens automatically. The procurement team is informed without having to open the email, locate the purchase order, or make a single manual update. What would typically take 10 to 15 minutes of manual processing is completed in seconds. ### Scenario 2: Confirmation with a Delivery Date Change This scenario covers a confirmation where the supplier cannot meet the original delivery date and is proposing a revised schedule. The details are otherwise correct, quantities and pricing match, but the date change needs to be reviewed before the order can be updated. #### Step 1: Confirmation Received and Parsed The supplier sends their purchase order confirmation via email. The Digital Worker picks it up immediately, reads through the email and attached document, and begins extracting and comparing all details against the purchase order in IFS Cloud. #### Step 2: Discrepancy Detected During the line-by-line comparison, the system identifies that the supplier has proposed a delivery date that differs from the original purchase order. Rather than posting the change automatically or ignoring it, the Digital Worker flags it as an exception requiring human review. A notification is sent to the procurement team via Microsoft Teams. The message includes full context: the original delivery date, the supplier's proposed date, and all other order details, so the reviewer has everything they need to make a decision without doing any additional investigation. #### Step 3: Human Decision Made The team member reviews the notification in Teams and approves the revised delivery date. This decision takes seconds rather than the time it would have taken to manually process the email, identify the discrepancy, determine the impact, update the system, and communicate back to the supplier. #### Step 4: ERP Updated and Supplier Notified The Digital Worker executes the approved change immediately. The purchase order in IFS Cloud is updated with the revised delivery date, with all other line item details unchanged. A confirmation is sent back to the supplier, and the team receives confirmation that the order has been updated. **Result:** The Digital Worker caught the delivery date change during validation, surfaced it to the right person with complete context, and updated the system the moment a decision was made. The procurement team stayed informed and in control without being buried in the processing work surrounding the decision. ## What These Scenarios Illustrate **Clean confirmations are fully automated.** When a supplier's confirmation matches the purchase order exactly, no human involvement is needed at any stage. The system handles extraction, comparison, ERP update, and team notification end to end. **Changes are caught before they reach the ERP.** Discrepancies in delivery dates, quantities, or pricing are identified during the comparison stage, before anything is written to IFS Cloud. This prevents incorrect data from entering the system and avoids the rework that would follow. **Exceptions come with full context attached.** When the system routes something for human review, it includes everything the reviewer needs to make a fast, informed decision: what changed, what the original value was, and what the impact is. Reviewers are making decisions, not investigating. **Approval triggers immediate action.** Once a decision is made in Teams, the Digital Worker executes it without delay. There is no separate step to update the system or notify the supplier, those happen automatically as part of the approval flow. **Outbound supplier communication is handled consistently.** Rather than individual team members drafting their own emails, confirmations and responses are generated by the system based on the purchase order context, ensuring professional, consistent communication at every touchpoint. # FAQ ### Capabilities & functionality **Q: What does Supplier Order Manager do for me?** A: It automatically manages supplier communications, updates purchase orders, handles delays & sends confirmations, saving you from repetitive email & data entry tasks. **Q: How quickly does Supplier Order Manager process supplier emails?** A: Updates are processed in near real-time. As soon as a supplier email arrives, Supplier Order Manager scans it, extracts key data & updates your ERP, often before you even see the message. **Q: How does Supplier Order Manager know what to update?** A: It parses supplier messages (emails, PDFs, attachments) to extract order details & cross-checks them with your ERP purchase orders. **Q: What if a supplier sends multiple changes in one email?** A: Supplier Order Manager can handle combined updates, such as a quantity change & a delivery date shift, applying them accurately & leaving a record of the original values. **Q: Does Supplier Order Manager work with international suppliers?** A: Yes. It supports multilingual email parsing & can handle multiple currencies, units of measure & tax configurations based on ERP data. ### Control & collaboration **Q: Will suppliers notice they're dealing with a digital worker?** A: No. Supplier Order Manager uses standard business communication templates, so messages look & feel like they're coming from your procurement team. **Q: Can I override Supplier Order Manager's decisions?** A: Yes. You can configure thresholds for automatic action & set rules for when human approval is required, such as price changes beyond a certain percentage. **Q: What happens if a supplier doesn't reply?** A: Supplier Order Manager automatically follows up with chaser emails. If no response is received after a defined period, it escalates the issue to a human user. **Q: How will this affect my role?** A: Instead of spending time updating spreadsheets & chasing emails, you'll focus more on supplier relationships, negotiations & strategic sourcing. ### Integration & technology **Q: How does Supplier Order Manager connect to ERP systems?** A: It uses standard APIs & projections to validate & update orders. **Q: Which APIs does Supplier Order Manager rely on?** A: Key APIs handle purchase order confirmations, updates with differences & document retrieval to ensure accurate, validated ERP updates. **Q: Can Supplier Order Manager integrate with tools beyond ERP?** A: Yes. It can connect with communication platforms such as Outlook, Gmail, & Teams to centralize supplier interactions. ### Security & reliability **Q: How are audit trails maintained?** A: Every action, including: email sent, order updated & confirmation logged, is written back to ERP history. Entries are stamped with the digital worker's ID, providing transparency & accountability. **Q: How does Supplier Order Manager handle data security?** A: Sensitive supplier & order data remains within your ERP & email environment. Supplier Order Manager uses secure API calls & logs all transactions for compliance. **Q: What happens if ERP is down?** A: Supplier Order Manager queues updates & retries once ERP connectivity is restored. Critical communications can still be drafted & logged, then reconciled once systems are back online. ### Customization & performance **Q: Can Supplier Order Manager adapt to our custom ERP workflows?** A: Yes. The technical design allows for configuration of custom rules, triggers, & validations to align with your ERP instance. **Q: How is performance measured?** A: Typical KPIs include the percentage of automated POs processed, processing time reduction, error rate & hours saved per week. These metrics are tracked through dashboards. # Case studies See how our customers are using this Digital Worker to transform their operations. {% embed url="" %} # How-to guides Choose a task below for clear, step-by-step instructions on the work you'll perform most.
Cover image
Navigate Learn the various components and their functions./files/443ppNpRgzNR2FsF2Yep/pages/F6QxLlyRAjQqAgJsPFCe
Add a toolEasily enhance and adapt functionality./files/YcLMbXMTSe2ABsraQ74q/pages/KXRoPPcLRp67m9ja0RJc
Debug a toolIdentify, review & diagnose bugs./files/ln4nAmHntdihyndj69CN/pages/tCJiNBcxcgKrbZu3Aj6w
# Navigate a Digital Worker The Agentic Platform has a library of digital workers. This guide will introduce you to the various components of a digital worker and their functions. ### Skills *Additional Skills can be added to your Digital Worker, enhancing their capabilities. These enhanced capabilities can enable your Digital Worker to perform a wider range of tasks and provide greater assistance.* {% stepper %} {% step %} Navigate to Skills. {% endstep %} {% step %} Click the skill to view the prompt & sub agent attributes. {% endstep %} {% step %} The Prompt field can be used to aid the Digital Worker when deciding which Skills to handoff to. {% endstep %} {% step %} Skills can also contain collections of tools. {% endstep %} {% endstepper %} ### Triggers *Triggers add actionability to a Digital Worker. Once a trigger has been added to a Digital Worker, it will observe and take action when the context is deemed to be relevant.* {% stepper %} {% step %} Navigate to Triggers. {% endstep %} {% step %} Depending on what connectors you have added to the system, any events coming from the connector side can act as a trigger. {% endstep %} {% step %} Multiple triggers can initiate the agent. {% endstep %} {% endstepper %} ### Settings *Here you can access and customize key aspects of the Digital Worker’s behavior. Each setting allows you to tailor how the Digital Worker executes tasks, collaborates with other Digital Workers and interacts with humans when needed.* {% stepper %} {% step %} **Human-in-the-Loop (HITL):** Configure escalation rules for scenarios where the system requires human review or encounters an error. 1. Supported notification channels: 1. Slack 2. Microsoft Teams 3. Email 4. Conditional routing (logic-based HITL flows) {% endstep %} {% step %} **Default Prompt:** Defines how other agents should interact with this agent when using it as a sub-agent. {% endstep %} {% step %} **Skill Attributes:** Specify the input variables the agent expects to receive when it is invoked by another agent. {% endstep %} {% step %} **Sub Agent Output Attributes:** Define the output variables that the agent will return after completing its task. {% endstep %} {% step %} **Enable Reflection:** Controls whether the agent performs self-reflection at each step of its plan. 1. Off: Default 2. On: When enabled, the agent pauses to evaluate it's progress, improving reliability at the cost of additional processing time. {% endstep %} {% step %} **Enable Concurrent Processing:** Allows the agent to handle multiple tasks in parallel, increasing throughput for high-volume workflows. {% endstep %} {% step %} **Agent Type:** Select the category to toggle between reasoning LLMS and non-reasoning LLMS. You can also set values for the levels of reasoning and verbosity. {% endstep %} {% endstepper %} ### Evaluation *This is a playground to test & debug the Digital Worker as we are developing, or at any time.* {% stepper %} {% step %} Conveniently test your Digital Worker at any time by clicking on the Evaluation button (arrow in circle) on the right sidebar. 1. For example: If a PDF normally submitted via email, you can upload a PDF instead and type out the relevant message in the chat option. {% endstep %} {% step %} While waiting for the agent to respond you can select Plan Flow (branching icon) to see the internal tool calls and sub agent handoffs. {% endstep %} {% step %} You can also see all your previous playground chats during evaluation by selecting Show Tabs (multi-line icon) on the right sidebar. {% endstep %} {% step %} Click the pencil icon next to any event in the list to view additional details. {% endstep %} {% endstepper %} # Add a tool ### Start a new tool {% stepper %} {% step %} Go to the **TOOLS** section. {% endstep %} {% step %} Enter a **TOOL NAME** and a **DESCRIPTION**. {% endstep %} {% step %} Click **SUBMIT**. {% endstep %} {% endstepper %} ### Configure the trigger & input form {% stepper %} {% step %} A trigger is automatically created and associated with a form. {% endstep %} {% step %} Edit the input form: 1. Example: *Create Ticket Form*. {% endstep %} {% step %} Add the necessary fields: 1. **Title** 2. **Description** 3. **Priority** (dropdown field) {% endstep %} {% step %} Click **SAVE**. {% endstep %} {% step %} Preview the form to see how it will look. {% endstep %} {% endstepper %} ### Add an output task {% stepper %} {% step %} Click on **+** → select **CUSTOM TASK**. {% endstep %} {% step %} Choose **LOOP TASK**. {% endstep %} {% step %} Search for and select "Widget Response Task**"**. {% endstep %} {% step %} Define the output format: 1. Add a **SUCCESS MESSAGE** (e.g., *Created Zendesk ticket*). 2. Include relevant fields to display: 1. **TICKET ID** → click **+** on the "Create Zendesk Ticket" tab → expand attributes → choose "Identity at Source**"** 2. **URL** → click **+** → select "URL**"** {% endstep %} {% step %} Click **SAVE**. {% endstep %} {% endstepper %} # Debug a tool ### Navigate to the tool {% stepper %} {% step %} Click on **TOOLS**. {% endstep %} {% step %} Select **DEBUG**. {% endstep %} {% step %} Choose the tool you want to debug (e.g., *Play-tok*). {% endstep %} {% endstepper %} ### Inspect event details {% stepper %} {% step %} In the execution view, you’ll see three tabs: 1. Execution 2. Input Attributes 3. Output Attributes {% endstep %} {% step %} At the **TRIGGER** confirm the input by clicking the checkmark. {% endstep %} {% step %} Review the **INTEGRATION EVENT**, which shows how the form was submitted. {% endstep %} {% step %} Look at the **DYNAMIC VALUES**, which display the output of the task. {% endstep %} {% endstepper %} ### Debug at the task level {% stepper %} {% step %} For an **API TASK**: 1. Check **COMPUTE ARGUMENTS** to see task inputs 2. Review the **STATUS** (done, failed, etc.) 3. Look at **DYNAMIC VALUES** for the output {% endstep %} {% step %} For a **WIDGET RESPONSE TASK**: 1. Again, check Compute Arguments, Status, and Dynamic Values. {% endstep %} {% endstepper %} ### Debug a failed execution {% stepper %} {% step %} Select a tool with a failed execution and go to **EVENTS**. {% endstep %} {% step %} Find and click on a failed entry. {% endstep %} {% step %} Check each task execution: 1. If the last task has nothing, the failure likely occurred at the API task level. 2. Review **INPUT ATTRIBUTES** for the trigger. 3. Review the **API TASK**: 1. Compute Arguments 2. Input Attributes 3. Status (likely failed) 4. Exception details (e.g., “unauthorized issue,” token expired). {% endstep %} {% step %} Click **OK** to confirm and continue checking other executions if needed. {% endstep %} {% endstepper %} # IFS Loops Agent Studio

IFS Loops Agent Studio Knowledge Base

**Welcome to the Agent Studio Knowledge Base** Everything you need to learn, build, and stay current with Agent Studio — all in one place.
Cover image
Essentials guideEverything you need to get up and running in Agent Studio — in under an hour. Follow along in sequence or jump to any topic to start building right away./pages/QmeYYk81iBdF8e7kTbjc/files/hr1tk8Gv0uIweIqrIt1x
Resource libraryA complete collection of articles and videos covering Agent Studio in depth. Return to it as a reference whenever you need more detail on any topic./pages/zKVb4lBDEBgv1ADV1ykY/files/o22oW5zles5gp76hQQa5
Release notesStay current with the latest updates, fixes, and feature releases in Agent Studio. Check back regularly to see what's new and what's changed./spaces/zpOPAqpE6CRlx7tqonUl/pages/O8A9Jf2Wf0j5tQ8qy7Ae/files/xD8PjiqhwAaGhEyvgKb5
IntegrationsConnect seamlessly with 65+ enterprise tools across CRM, ticketing & more/pages/DBHws38eO8AG9o0lUs7k/files/pmFl9ohNMNkdeS7NxWmG
# Essentials guide Everything you need to get up and running in Agent Studio. Follow along in sequence or jump to any topic to start building right away.
Cover image
1- Introduction to Agent StudioWhat Agent Studio is and how it is structured./files/U7caKinbqsUoAx0cA5xW/pages/Epcz9WGPlSIfC0dq7t0x
2. Workspaces and EnvironmentsHow workspaces and UAT vs Production environments function./files/Vk1u6mJtNbptvHSRey4H/pages/mCcp1bsgNxCp1kd2zp0t
3. Installing a Digital WorkerMarketplace, Local Template Store, and the import wizard./files/gexUqIQVIkX8Db10PafI/pages/EawrY8dwGWvVRF1N4p4w
4. Customizing a Digital WorkerChanging instructions, adding skills & adding tools./files/YITkXZWVl35YDtRQNBjp/pages/AXDwaeLEeBaIqfQsZxJl
5. Configuring Triggers & Human-in-the-LoopTriggers, Human-in-the-Loop, structured output, and API access/files/4PoTeqKI5JHMQbihrAwu/pages/SuTVibMgCFMcrfcobNQq
6. Using the MetaPrompterEfficient & conversational customization./files/UzUqlsXuC5nSdFu0b91h/pages/ESUS4e5IMoZYjkNKQIqD
7. Evaluating a Digital WorkerScorecards, rubrics, evaluation results, and dashboards./files/e9G0WmboAYnrJ01Xmr0B/pages/FH6fkhzFxQTY2un4YfqO
8. Versioning & Deploying to ProductionDraft vs versioned copies, publishing, and pushing to production./files/tjReNFedeajLII4EZjsS/pages/aLK7kTyUUndnEmisePRn
# 1. Introduction to Agent Studio {% embed url="" %} *What Agent Studio is and how it is structured.* ## Watch these Agent Studio videos to learn more: 1. **Introduction to Agent Studio** 2. [Workspaces and Environments](/agentstudio/essentials-guide/2.-workspaces-and-environments) 3. [ Installing a Digital Worker](/agentstudio/essentials-guide/3.-installing-a-digital-worker) 4. [Customizing a Digital Worker](/agentstudio/essentials-guide/4.-customizing-a-digital-worker) 5. [Configuring Triggers & Human-in-the-Loop](/agentstudio/essentials-guide/5.-configuring-triggers-and-human-in-the-loop) 6. [Using the MetaPrompter](/agentstudio/essentials-guide/6.-using-the-metaprompter) 7. [Evaluating a Digital Worker](/agentstudio/essentials-guide/7.-evaluating-a-digital-worker) 8. [Versioning & Deploying to Production](/agentstudio/essentials-guide/8.-versioning-and-deploying-to-production) # 2. Workspaces and Environments {% embed url="" %} *How workspaces and UAT vs Production environments function.* ## Watch these Agent Studio videos to learn more: 1. [Introduction to Agent Studio](/agentstudio/essentials-guide/1.-introduction-to-agent-studio) 2. **Workspaces and Environments** 3. [ Installing a Digital Worker](/agentstudio/essentials-guide/3.-installing-a-digital-worker) 4. [Customizing a Digital Worker](/agentstudio/essentials-guide/4.-customizing-a-digital-worker) 5. [Configuring Triggers & Human-in-the-Loop](/agentstudio/essentials-guide/5.-configuring-triggers-and-human-in-the-loop) 6. [Using the MetaPrompter](/agentstudio/essentials-guide/6.-using-the-metaprompter) 7. [Evaluating a Digital Worker](/agentstudio/essentials-guide/7.-evaluating-a-digital-worker) 8. [Versioning & Deploying to Production](/agentstudio/essentials-guide/8.-versioning-and-deploying-to-production) # 3. Installing a Digital Worker *Marketplace, Local Template Store, and the import wizard.* {% embed url="" %} ## Watch these Agentic Studio videos to learn more: 1. [Introduction to Agent Studio](/agentstudio/essentials-guide/1.-introduction-to-agent-studio) 2. [Workspaces and Environments](/agentstudio/essentials-guide/2.-workspaces-and-environments) 3. **Installing a Digital Worker** 4. [Customizing a Digital Worker](/agentstudio/essentials-guide/4.-customizing-a-digital-worker) 5. [Configuring Triggers & Human-in-the-Loop](/agentstudio/essentials-guide/5.-configuring-triggers-and-human-in-the-loop) 6. [Using the MetaPrompter](/agentstudio/essentials-guide/6.-using-the-metaprompter) 7. [Evaluating a Digital Worker](/agentstudio/essentials-guide/7.-evaluating-a-digital-worker) 8. [Versioning & Deploying to Production](/agentstudio/essentials-guide/8.-versioning-and-deploying-to-production) # 4. Customizing a Digital Worker *Changing instructions, adding skills & adding tools.* {% embed url="" %} ## Watch these Agent Studio videos to learn more: 1. [Introduction to Agent Studio](/agentstudio/essentials-guide/1.-introduction-to-agent-studio) 2. [Workspaces and Environments](/agentstudio/essentials-guide/2.-workspaces-and-environments) 3. [ Installing a Digital Worker](/agentstudio/essentials-guide/3.-installing-a-digital-worker) 4. **Customizing a Digital Worker** 5. [Configuring Triggers & Human-in-the-Loop](/agentstudio/essentials-guide/5.-configuring-triggers-and-human-in-the-loop) 6. [Using the MetaPrompter](/agentstudio/essentials-guide/6.-using-the-metaprompter) 7. [Evaluating a Digital Worker](/agentstudio/essentials-guide/7.-evaluating-a-digital-worker) 8. [Versioning & Deploying to Production](/agentstudio/essentials-guide/8.-versioning-and-deploying-to-production) # 5. Configuring Triggers & Human-in-the-Loop *Triggers, Human-in-the-Loop, structured output, and API access* {% embed url="" %} ## Watch these Agent Studio videos to learn more: 1. [Introduction to Agent Studio](/agentstudio/essentials-guide/1.-introduction-to-agent-studio) 2. [Workspaces and Environments](/agentstudio/essentials-guide/2.-workspaces-and-environments) 3. [ Installing a Digital Worker](/agentstudio/essentials-guide/3.-installing-a-digital-worker) 4. [Customizing a Digital Worker](/agentstudio/essentials-guide/4.-customizing-a-digital-worker) 5. **Configuring Triggers & Human-in-the-Loop** 6. [Using the MetaPrompter](/agentstudio/essentials-guide/6.-using-the-metaprompter) 7. [Evaluating a Digital Worker](/agentstudio/essentials-guide/7.-evaluating-a-digital-worker) 8. [Versioning & Deploying to Production](/agentstudio/essentials-guide/8.-versioning-and-deploying-to-production) # 6. Using the MetaPrompter {% embed url="" %} *Efficient & conversational customization.* ## Watch these Agent Studio videos to learn more: 1. [Introduction to Agent Studio](/agentstudio/essentials-guide/1.-introduction-to-agent-studio) 2. [Workspaces and Environments](/agentstudio/essentials-guide/2.-workspaces-and-environments) 3. [ Installing a Digital Worker](/agentstudio/essentials-guide/3.-installing-a-digital-worker) 4. [Customizing a Digital Worker](/agentstudio/essentials-guide/4.-customizing-a-digital-worker) 5. [Configuring Triggers & Human-in-the-Loop](/agentstudio/essentials-guide/5.-configuring-triggers-and-human-in-the-loop) 6. **Using the MetaPrompter** 7. [Evaluating a Digital Worker](/agentstudio/essentials-guide/7.-evaluating-a-digital-worker) 8. [Versioning & Deploying to Production](/agentstudio/essentials-guide/8.-versioning-and-deploying-to-production) # 7. Evaluating a Digital Worker {% embed url="" %} *Scorecards, rubrics, evaluation results, and dashboards.* ## Watch these Agent Studio videos to learn more: 1. [Introduction to Agent Studio](/agentstudio/essentials-guide/1.-introduction-to-agent-studio) 2. [Workspaces and Environments](/agentstudio/essentials-guide/2.-workspaces-and-environments) 3. [ Installing a Digital Worker](/agentstudio/essentials-guide/3.-installing-a-digital-worker) 4. [Customizing a Digital Worker](/agentstudio/essentials-guide/4.-customizing-a-digital-worker) 5. [Configuring Triggers & Human-in-the-Loop](/agentstudio/essentials-guide/5.-configuring-triggers-and-human-in-the-loop) 6. [Using the MetaPrompter](/agentstudio/essentials-guide/6.-using-the-metaprompter) 7. **Evaluating a Digital Worker** 8. [Versioning & Deploying to Production](/agentstudio/essentials-guide/8.-versioning-and-deploying-to-production) # 8. Versioning & Deploying to Production {% embed url="" %} *Draft vs versioned copies, publishing, and pushing to production.* ## Watch these Agent Studio videos to learn more: 1. [Introduction to Agentic Studio](/agentstudio/essentials-guide/1.-introduction-to-agent-studio) 2. [Workspaces and Environments](/agentstudio/essentials-guide/2.-workspaces-and-environments) 3. [ Installing a Digital Worker](/agentstudio/essentials-guide/3.-installing-a-digital-worker) 4. [Customizing a Digital Worker](/agentstudio/essentials-guide/4.-customizing-a-digital-worker) 5. [Configuring Triggers & Human-in-the-Loop](/agentstudio/essentials-guide/5.-configuring-triggers-and-human-in-the-loop) 6. [Using the MetaPrompter](/agentstudio/essentials-guide/6.-using-the-metaprompter) 7. [Evaluating a Digital Worker](/agentstudio/essentials-guide/7.-evaluating-a-digital-worker) 8. **Versioning & Deploying to Production** # Resource library A complete collection of articles and videos covering Agent Studio in depth. Return to it as a reference whenever you need more detail on any topic.
Getting Started [*What Is Agent Studio?*](/agentstudio/resource-library/explainers/what-is-agent-studio) Platform overview, four pillars, building blocks, who it's for [*Workspaces and Environments Explained*](/agentstudio/resource-library/explainers/workspaces-and-environments-explained) Isolation model, UAT vs Production, navigation, key rules [*Understanding the Agent Development Life Cycle (ADLC)*](/agentstudio/resource-library/explainers/understanding-the-agent-development-life-cycle-adlc) 5-stage ADLC, versioning rules, what can change in Production
Best Practices [*Versioning: Drafts, Versions, and Promotion*](/agentstudio/resource-library/explainers/versioning-drafts-versions-and-promotion) Draft vs versioned copies, format, Marketplace installs, immutability [*Customizing Digital Workers: Instructions, Skills, Tools, and Triggers* ](/agentstudio/resource-library/explainers/customizing-digital-workers-instructions-skills-tools-and-triggers)Instructions, skills, tools, triggers, HITL, structured output, concurrency [*Evaluating Digital Workers: Scorecards, Rubrics, and Dashboards*](/agentstudio/resource-library/explainers/evaluating-digital-workers-scorecards-rubrics-and-dashboards) Scorecards, their purpose, how rubrics and dashboards function
MetaPrompter [*The Meta Prompter: Customizing with Plain Language*](/agentstudio/resource-library/explainers/the-metaprompter-customizing-with-plain-language) Plain-language config, how it works, availability, typical use
How to Guides [*How to Find and Use the Marketplace*](/agentstudio/resource-library/how-to-guides/find-and-use-the-marketplace) How to Use the Marketplace Import wizard, connector mapping, local template store [*How to Customize Instructions*](/agentstudio/resource-library/how-to-guides/customize-instructions) When and where to change, writing tips, MetaPrompter, after changes [*How to Manage Workspaces and Environments*](/agentstudio/resource-library/how-to-guides/manage-workspaces-and-environments) Create, navigate, switch environments, planning tips [*How to Set Up Scorecards*](/agentstudio/resource-library/how-to-guides/set-up-scorecards) Scorecards, rubrics, publishing, interaction results, dashboards
# Explainers Get started with the Explainers to broaden your knowledge on Agentic Studio, from understanding the platform's core structure to configuring, versioning, evaluating your digital worker
What Is Agentic Studio? What is Agentic Studio? Platform overview, four pillars, \ building blocks, who it's for [*What Is Agentic Studio?*](https://app.gitbook.com/o/KhWmVAIJ02ASl3wYnaK7/s/RIbp8Kp04VLXld01gmAs/~/edit/~/changes/144/resource-library/explainers/what-is-agentic-studio)
Workspaces and Environments Explained Isolation model, UAT vs Production, navigation, key rules [*Workspaces and Environments Explained*](https://app.gitbook.com/o/KhWmVAIJ02ASl3wYnaK7/s/RIbp8Kp04VLXld01gmAs/~/edit/~/changes/144/resource-library/explainers/workspaces-and-environments-explained)
Understanding the Agent Development Life Cycle (ADLC) 5-stage ADLC, versioning rules, what can change in Production [*Understanding the Agent Development Life Cycle (ADLC)*](https://app.gitbook.com/o/KhWmVAIJ02ASl3wYnaK7/s/RIbp8Kp04VLXld01gmAs/~/edit/~/changes/144/resource-library/explainers/understanding-the-agent-development-life-cycle-adlc)
Versioning: Drafts, Versions, and Promotion Draft vs versioned copies, format, Marketplace installs, immutability [*Versioning: Drafts, Versions, and Promotion*](https://app.gitbook.com/o/KhWmVAIJ02ASl3wYnaK7/s/RIbp8Kp04VLXld01gmAs/~/edit/~/changes/144/resource-library/explainers/versioning-drafts-versions-and-promotion)
Customizing Digital Workers: Instructions, Skills, Tools, and Triggers Instructions, skills, tools, triggers, HITL, structured output, concurrency [*Customizing Digital Workers: Instructions, Skills, Tools, and Triggers*](https://app.gitbook.com/o/KhWmVAIJ02ASl3wYnaK7/s/RIbp8Kp04VLXld01gmAs/~/edit/~/changes/144/resource-library/explainers/customizing-digital-workers-instructions-skills-tools-and-triggers)
The Meta Prompter: Customizing with Plain Language Plain-language config, how it works, availability, typical use [*The Meta Prompter: Customizing with Plain Language*](https://app.gitbook.com/o/KhWmVAIJ02ASl3wYnaK7/s/RIbp8Kp04VLXld01gmAs/~/edit/~/changes/144/resource-library/explainers/the-meta-prompter-customizing-with-plain-language)
Evaluating Digital Workers: Scorecards, Rubrics, and Dashboards Scorecards, their purpose, how rubics and dashboards function [*Evaluating Digital Workers: Scorecards, Rubrics, and Dashboards*](https://app.gitbook.com/o/KhWmVAIJ02ASl3wYnaK7/s/RIbp8Kp04VLXld01gmAs/~/edit/~/changes/144/resource-library/explainers/evaluating-digital-workers-scorecards-rubrics-and-dashboards)
# What Is Agent Studio? Agent Studio is IFS Loops' central platform for building, deploying, and managing digital workers. Whether you are installing a pre-built template from the Marketplace or assembling a custom worker from scratch, Agent Studio provides the environment, tooling, and lifecycle management to take a digital worker from concept to production. Think of it as a control tower for automation: one place where teams configure how digital workers behave, test them safely before go-live, monitor their performance in real time, and continuously improve their outputs over time. ### **The Four Pillars of This Release** The current release of Agent Studio is built around four core areas: * Marketplace: a store of pre-packaged digital worker and skill templates, maintained and updated by the IFS Loops team. * Agent Studio: the full development and management environment for digital workers, including installation, customization, testing, and monitoring. * FSM Digital Workers: purpose-built digital workers for Field Service Management workflows. * Connectors: integrations with IFS and related third-party applications that give digital workers access to real system data and actions. ### **Who Uses Agent Studio?** Agent Studio is designed for enablement teams, solution architects, and business administrators responsible for deploying digital workers across an organization. You do not need to be a developer to get started, as many configuration tasks use a conversational interface. A working understanding of the product and processes you are automating will help you get the most from the platform. ### **The Core Building Blocks** {% tabs %} {% tab title="Workspaces" %} **Workspaces** Workspaces let you segment digital workers by business area, site, or use case. Each workspace keeps its data, context, and learning separate, so insights from one part of the business do not bleed into another. For example, a manufacturer running two distinct sites can create two workspaces, one per site, to keep operations cleanly isolated. Up to five workspaces are supported per tenant. {% endtab %} {% tab title="Environments" %} **Environments: UAT and Production** Every workspace includes two environments. UAT (User Acceptance Testing) is where you make changes and validate behavior before anything reaches live operations. Production is where the digital worker handles real events. Changes always start in UAT and are promoted to Production only once they have been tested and published. {% endtab %} {% tab title="Digital Workers " %} **Digital Workers** A digital worker is the core unit of automation in Agent Studio. It combines instructions (what to do and how to reason), skills (modular capabilities it can invoke), tools (actions it can take in external systems), and triggers (events that start it automatically) into a single, managed entity. {% endtab %} {% tab title="Skills and Tools " %} **Skills and Tools** Skills are reusable, modular capabilities that a digital worker can call when it detects the right intent. Tools give a digital worker the ability to take real actions, such as querying systems, creating records, and sending messages, by connecting to your business applications through connectors. {% endtab %} {% endtabs %} ### **The Agent Development Life Cycle (ADLC)** Agent Studio structures digital worker delivery through a five-stage lifecycle that mirrors software development practices: | Stage | What Happens | | --------- | -------------------------------------------------------------------------------------- | | Configure | Install or modify a digital worker: instructions, skills, tools, and triggers. | | Test | Run sanity tests and evaluate outputs in the UAT environment. | | Deploy | Publish a versioned copy and promote it to Production. | | Monitor | Track usage, events, and performance in real time. | | Learn | Review evaluation scores, identify gaps, and iterate on instructions or configuration. | ### **Why This Matters** Traditional automation tools treat processes as static. Agent Studio is built around the reality that processes change: systems are updated, requirements shift, and edge cases emerge. The ADLC, paired with versioning and evaluation tooling, gives teams the confidence to make those changes safely and predictably, without disrupting live operations. [*Tip: Digital workers imported from the Marketplace are typically about 60% ready out of the box. Agent Studio provides all the tools needed to close the remaining gap for your organization.* ](#user-content-fn-1)[^1] [^1]: # Workspaces and Environments Explained Agent Studio uses a two-level structure of workspaces and environments to give you precise control over where Digital Workers operate and what data they can access. Understanding this structure is essential before you install or configure anything. #### Workspaces A workspace is an isolated container for a specific business area, site, or use case. Each workspace maintains its own data, context, and learning independently, which means a Digital Worker in one workspace cannot be influenced by activity in another. Connectors are also configured per workspace, so each workspace connects to its own external systems independently. Workspaces support three separation models. You choose based on your organizational needs: * **Environment separation:** handled automatically within each workspace. Every workspace includes UAT and Production environments by default. * **Functional separation:** create one workspace per business function, such as one for Accounts Payable and another for Customer Order Management, each with its own Digital Workers, connectors, and data. * **Business unit separation:** create workspaces per division or region, each fully isolated from the others. This isolation is particularly useful for organizations running multiple sites or business units with distinct processes. For example, a field service team with operations in two countries can create one workspace per region, ensuring local SLA rules, terminology, and system configurations stay separate. #### Planning your workspace structure For most use cases, the right number of workspaces maps directly to the number of distinct functional areas or business units you need to keep separate. Two Digital Workers serving different functions, such as AP and COM, typically warrant two workspaces. You would only need additional workspaces if you require further separation, such as regional deployments or a dedicated sandbox for experimentation. {% hint style="info" %} **Limit:** Each tenant can create a maximum of five workspaces. Plan your workspace structure carefully before you begin. {% endhint %} #### Environments Every workspace includes exactly two environments: * **UAT (User Acceptance Testing):** a safe space to install, modify, and test Digital Workers before they handle real operations. All changes start here. * **Production:** the live environment where Digital Workers process real events. Digital Workers are promoted to Production only after being tested and published in UAT. # Understanding the Agent Development Life Cycle (ADLC) The Agent Development Life Cycle, or ADLC, is the structured process Agent Studio uses to manage how digital workers are built, validated, deployed, and improved over time. It is modeled on software development best practices, adapted for the unique requirements of AI-powered automation. Following the ADLC reduces risk, increases confidence in production deployments, and gives teams a repeatable way to manage change. ### **The Five Stages** #### **1. Configure** This is where you set up or update a digital worker. Configuration includes setting instructions (the policy the worker follows), adding or modifying skills (what capabilities it can invoke), connecting tools (what actions it can take), defining triggers (what events start it automatically), and configuring HITL rules (when a human needs to review or approve a step). All configuration happens on the Draft version of a digital worker in the UAT environment. #### **2. Test** Before promoting changes to Production, you validate the digital worker's behavior in UAT. This includes running sanity tests to confirm that instructions, skills, and tools behave as expected, reviewing outputs for accuracy and format, and checking evaluation scores against your rubric criteria. #### **3. Deploy** When testing is complete, you publish the Draft as a new versioned copy (for example, version 1.5 becomes 1.6), then promote that version to Production. Versioned copies are read-only and cannot be changed after publishing, which means you always have a stable, auditable record of what is running in Production. #### **4. Monitor** Agent Studio provides real-time observability into digital workers running in Production. You can review individual events and interactions, track usage over time, and identify anomalies or failures as they occur. #### **5. Learn** The evaluation framework lets you score interactions against defined rubrics and scorecards. Aggregated results roll up into dashboards that reveal trends, highlight failure modes, and surface where instructions, skills, or tools need adjustment. Insights from this stage feed directly back into the Configure stage, creating a continuous improvement loop. ### **Versioning in the ADLC** Versioning is what makes the ADLC safe and auditable. Every digital worker starts with a Draft. When you publish, the Draft becomes a versioned copy using a major.minor format, for example 1.4. Publishing always increments the minor version. The Draft continues to exist as the starting point for the next iteration. #### Key version rules: * You can only edit the Draft. Published versions are read-only. * You can promote any version to Production, not just the most recent one. * The platform does not enforce which version is pushed or whether it has been tested. Operational discipline and an internal release process are required to manage this safely. {% hint style="info" %} *Before pushing any version to Production, document your UAT test evidence, get an approval from the relevant owner, and confirm you have a rollback plan in place.* {% endhint %} ### **What Can Be Changed in Production** Once a digital worker is running in Production, it is not editable. Three settings can be adjusted without republishing: * HITL configuration: update escalation rules or routing channels without creating a new version. * Model parameters: tune the underlying model settings for performance. * Parallelism and concurrency: adjust how many tasks the worker can process simultaneously. For any other change, return to UAT, update the Draft, publish a new version, and promote it to Production. # Versioning: Drafts, Versions, and Promotion Versioning in Agent Studio is the mechanism that keeps digital worker changes controlled, auditable, and reversible. It ensures that Production environments always run a known, stable configuration while teams are free to iterate in UAT. ### **Draft vs Versioned Copies** Every digital worker has exactly one Draft at any time. The Draft lives in UAT and is the only copy you can actively edit. When your changes are complete and testing is done, you publish the Draft. At that point, a versioned copy is created and the Draft resets as the starting point for your next round of changes. Versioned copies are read-only. Once published, they cannot be modified. This immutability is what makes the version history reliable and auditable. ### **Version Format** Versions use a major. Minor format, for example 1.4 or 2.0. Publishing always increments the minor version. Major version changes are typically reserved for significant structural changes to the digital worker. | Version Event | Result | | -------------------------------------- | ------------------------------------------------------------- | | Install from Marketplace (version 1.5) | Digital worker is set to 1.5; a Draft copy of 1.5 is created. | | Make changes and publish | Draft becomes version 1.6; a new Draft is created from 1.6. | | Make further changes and publish | Draft becomes version 1.7; and so on. | | Push version 1.6 to Production | Production runs 1.6 even if 1.7 exists in UAT. | ### **Installing from the Marketplace** When you install a digital worker from the Marketplace, the template version is preserved exactly. If the Marketplace template is at version 1.5, your installed digital worker starts at version 1.5, and a Draft copy of 1.5 is created as your starting point for customization. ### **Pushing to Production** You can promote any published version, not just the most recent one, to Production. This gives you the flexibility to roll back to an earlier stable version if a newer one introduces issues. Important: The platform does not enforce version order or require evidence of testing before promotion. Your team is responsible for defining and following a release process that ensures only tested, approved versions reach Production. ### **Why Immutable Versions Matter** Immutable versions mean that what ran in Production on a given date is always recoverable and comparable. This is critical for audits, incident investigation, and evaluating whether a change to instructions or skills actually improved performance. If you need to know exactly what configuration was running last Tuesday, you can look it up by version number. # Customizing Digital Workers: Instructions, Skills, Tools, and Triggers Digital worker templates from the Marketplace are a strong starting point, but most organizations need to customize them to match their specific processes, terminology, and systems. Agent Studio provides several layers of customization, from simple instruction changes to adding new capabilities and integrations. Templates are typically around 60% applicable out of the box. The customization layer is where you close the remaining gap. ### **Changing Instructions** Instructions are the core policy of a digital worker. They define what the worker does, how it reasons, what format it returns, and how it handles edge cases. Because instructions live in the prompt, even small wording changes can reliably shift the worker's downstream behavior without requiring any changes to skills or tools. For example, if a digital worker currently produces a free-form summary, you can update its instructions to require a structured JSON output with specific required fields, null-handling rules, and targeted follow-up questions for missing data. The same inbound event will now produce a machine-readable result instead of a narrative response. {% hint style="info" %} *Start with instruction changes before adding new skills or tools. Often, adjusting how the worker reasons or formats its output is all that is needed.* {% endhint %} ### **Adding Skills** Skills are modular capabilities that a digital worker invokes when it detects the right intent. Rather than embedding every behavior in the instructions, you can keep instructions lean and extend the worker's capabilities through targeted skills. When defining a skill, the most important configuration is the invocation parameters, specifically the prompts or intent patterns that should activate the skill. Well-defined invocation criteria ensure the skill is called reliably when needed, and not called when it should not be. #### Example skills and their typical triggers: * Get Work Order Details: triggers on requests such as "what is the status of WO-12345?" * Search Knowledge Base: triggers on questions such as "how do I reset the controller?" * Create Dispatch Recommendation: triggers on requests such as "who should we send?" If the worker cannot find the required information to call a skill, it should ask targeted follow-up questions rather than proceeding with incomplete data. ### **Adding Tools** Tools give a digital worker real execution power, meaning the ability to take actions beyond reasoning and text generation. In Agent Studio, tools can come from three sources: * Connector tools: API operations imported from a connected system such as IFS, ServiceNow, or a third-party application. * Out-of-the-box tools: pre-built capabilities provided by IFS Loops. * Custom tools: tools you build to match specific organizational processes. Depending on the connector, tools may be exposed as traditional APIs or as MCP-based capabilities, allowing the digital worker to reliably discover and invoke the right function at the right time. ### **Adding Triggers** Triggers start a digital worker automatically in response to an external event from a connected system. The event payload carries key context, including IDs, timestamps, and changed fields, which the worker uses to determine its next step. #### Example triggers: * Service Work Order Created (IFS): sends the work order ID and key fields. The digital worker auto-classifies priority, flags missing information, and recommends next steps. * Incident Status Changed (ServiceNow): on escalation, sends incident context. The digital worker summarizes impact, retrieves relevant knowledge base articles, and drafts the escalation update. ### **Configuring Human-in-the-Loop (HITL)** * HITL adds an approval or decision step to a digital worker's configuration so that sensitive, ambiguous, or high-impact actions are reviewed by a person before the execution continues. * HITL can be routed to Microsoft Teams, Slack, or other collaboration channels. Configuration adapts to the chosen channel. In Slack, for example, you select the target workspace and channel where requests should be posted. * HITL can also be made conditional, triggering only when specific rules are met, such as low confidence, missing required fields, an exception condition, or a high-value transaction, while keeping the rest of the process fully automated. This allows different humans or teams to handle different decision points within the same configuration. ### **Structured Output** Structured output ensures a digital worker returns results in a predictable schema, for example a named JSON object, so that downstream systems, dashboards, or integrations can reliably parse and use the response. To configure structured output, define an explicit output contract that maps each required field to its source in the conversation or tool results. Specify required versus optional fields, allowed values, and what to do when data is missing (for example, set the field to null and ask a targeted follow-up question). Adding this contract to the digital worker's instructions makes outputs consistent and integration-ready. ### **Concurrency** Concurrency controls how many parallel tasks a digital worker can run simultaneously. This matters particularly for high-volume scenarios where instructions like "for each entity, do this" can trigger a fan-out into many parallel executions. Unbounded fan-out can significantly increase token usage and drive spikes in system resource consumption. Setting a concurrency limit caps parallel fan-outs, paces execution, and keeps the process predictable while still benefiting from parallelism where it matters. # The MetaPrompter: Customizing with Plain Language The MetaPrompter is Agent Studio's conversational interface for modifying digital workers. Instead of manually editing instructions, skills, or configuration fields, you describe what you want to change in plain language. The MetaPrompter generates a plan, applies the changes to the Draft, and leaves you in control of when to publish and promote. It is especially useful for teams working with Marketplace templates that need targeted adjustments to align with specific organizational processes. #### What You Can Do with MetaPrompter * Update a digital worker's instructions and output format. * Adjust how the worker handles edge cases or ambiguous inputs. * Refine prompts, reasoning steps, and behavior to better match your operations. ### **How It Works** The MetaPrompter uses a simple two-step interaction: you describe a change, it proposes a plan, you confirm, and it applies the update to the Draft. The final publish and promotion steps remain under your control. 1. Launch the MetaPrompter chat from the digital worker's Graph view (accessible from the Home page or the Compose tab). 2. Type the change you want to make in plain language. 3. Review the plan the MetaPrompter proposes. It will outline exactly what it intends to update. 4. Confirm the plan. The MetaPrompter applies the changes to the Draft copy of the digital worker. 5. Run sanity tests to validate the change behaves as expected. 6. Publish a new version and promote it to Production when you are ready. ### **Availability and Constraints** MetaPrompter is only available for the Draft version of a digital worker in the UAT environment. It cannot be used on published versions or in Production. #### **Typical Use** The most common use of MetaPrompter is immediately after installing a Marketplace template. The template provides a solid foundation, but it often needs refinement for your organization's specific terminology, escalation rules, output formats, or edge case handling. MetaPrompter lets you make those refinements conversationally, without needing to understand every underlying technical detail of the digital worker's configuration # Evaluating Digital Workers: Scorecards, Rubrics, and Dashboards Agent Studio includes a built-in evaluation framework that scores every digital worker interaction after it closes. This framework lets you define what good looks like, measure whether the digital worker is meeting that standard, and identify where instructions, skills, or tools need adjustment. The evaluation approach mirrors Auto-QA: structured scorecards and rubrics produce consistent, auditable scores that roll up into dashboards for trend monitoring. ### **Scorecards** A scorecard defines what gets measured and how each dimension is weighted. It is the top-level structure that ties your evaluation criteria together. A typical scorecard includes: * A defined purpose, for example Support Quality, Dispatch Triage Quality, or Sales Qualification. * Score categories (high-level measurement buckets) such as Policy and Process Adherence, Accuracy and Resolution Quality, Communication Quality, or Safety and Compliance. * Weights for each category, for example 40% accuracy, 30% process adherence, 20% communication, 10% compliance. * One or more rubric criteria attached to each category. Once configured, a scorecard is published to one or more digital workers in a specific environment. Start with UAT before publishing to Production. ### **Rubrics** Rubrics define the specific criteria used to score an interaction. Each rubric item is a clear, observable check or question that a reviewer, human or automated, can answer consistently. Examples of rubric criteria: * Did the digital worker follow the required verification steps? * Was the final answer accurate and complete? * Did the digital worker use the correct tool or configuration? * Was the tone professional and helpful? For each criterion, you set a scoring scale (for example, Pass/Fail, 1 to 5, or 0 to 2) and add guidance for reviewers: what good looks like, common failure modes, and what evidence to look for in the conversation transcript. Rubric versions should be maintained over time so that historical interactions remain comparable and auditable even as criteria evolve. ### **Interaction-Level Results** After a conversation closes, it is scored automatically. The rubric evaluation is displayed directly on the conversation view. * Open any agent interaction to see criterion-level scores, category rollups, and the overall score. * Drill into failures to understand which rubric criteria drove the score down and what evidence in the conversation supports the rating. * Use the interaction view for spot-checking, incident review, and debugging specific changes to instructions, skills, or tools. ### **Evaluation Dashboards** Dashboards aggregate scores across interactions so you can monitor quality at scale and detect regressions over time. Dashboard capabilities include: * Trend tracking: overall score trends by day or week. * Slicing: filter by digital worker, workspace, environment, channel, score category, or individual rubric criterion. * Distribution monitoring: percentage of interactions below a threshold, pass/fail rates, and category breakdowns. * Release validation: compare scores before and after a new version is promoted to Production. Use dashboards not just to track averages, but to understand distributions. A high average score that hides a tail of very low-scoring interactions is a sign of a systematic failure mode, not overall good performance. # How-to-Guides The how-to guides provide step-by-step instructions for importing from the Marketplace, editing instructions, managing workspaces, and setting up scorecards.
Find and Use the Marketplace How to Use the Marketplace Import wizard, connector mapping, local template store [*How to Find and Use the Marketplace*](https://app.gitbook.com/o/KhWmVAIJ02ASl3wYnaK7/s/RIbp8Kp04VLXld01gmAs/~/edit/~/changes/149/resource-library/how-to-guides/how-to-find-and-use-the-marketplace)
Customize a Digital Worker's Instructions When and where to change, writing tips, MetaPrompter, after changes [*How to Customize Instructions*](https://app.gitbook.com/o/KhWmVAIJ02ASl3wYnaK7/s/RIbp8Kp04VLXld01gmAs/~/edit/~/changes/149/resource-library/how-to-guides/how-to-customize-instructions)
Create and Manage a Workspace Create, navigate, switch environments, planning tips [*How to Manage Workspaces*](https://app.gitbook.com/o/KhWmVAIJ02ASl3wYnaK7/s/RIbp8Kp04VLXld01gmAs/~/edit/~/changes/149/resource-library/how-to-guides/how-to-manage-workspaces)
Set Up and Use Evaluation Scorecards Scorecards, rubrics, publishing, interaction results, dashboards [*How to Set Up Scorecards*](https://app.gitbook.com/o/KhWmVAIJ02ASl3wYnaK7/s/RIbp8Kp04VLXld01gmAs/~/edit/~/changes/149/resource-library/how-to-guides/how-to-set-up-scorecards)
# Find and Use the Marketplace The Marketplace is the central store for pre-packaged digital worker and skill templates maintained by the IFS Loops team. This guide walks you through how to find a template and import it into your workspace. **Before You Start** * You must be logged in to Agent Studio with appropriate permissions. * You must have a workspace created and be working in the UAT environment. * Any connectors required by the template you intend to import should be available in your environment. **Steps: Import a Digital Worker from the Marketplace** 1. Log in to Agent Studio. 2. Navigate to the Digital Workers page from the main navigation. 3. Click the Create Digital Worker button. 4. Select Import From Template. 5. On the templates page, you will see two tabs: Local Templates (published by users in your organization) and Marketplace (published by IFS Loops). Select Marketplace. 6. Browse or search for the template you need. Select it to view details. 7. Follow the import wizard. You will be asked to map the connectors required by this digital worker to your available connectors. 8. Once connectors are mapped, click Import. The digital worker is installed in your workspace in the UAT environment, at the version available in the Marketplace. 9. The digital worker now appears on your Digital Workers page. A Draft copy is automatically created as your starting point for customization. *Note: Digital workers can only be installed into the UAT environment. They cannot be deployed directly to Production.* **What Happens After Import** Once the digital worker is installed, you can open it from the Digital Workers page to access the Home, Compose, and Evaluate tabs. From there, you can review the digital worker's default configuration, use the Meta Prompter to make adjustments, run tests, and when ready, publish and promote to Production. **Using the Local Template Store** In addition to the Marketplace, your organization can maintain a local template store: a shared library of modified digital workers and skills that your team has customized and published internally. These appear on the Local Templates tab of the template selection screen. Publishing to the local template store allows configurations tailored to your organization's processes to be reused across workspaces without starting from scratch each time. # Manage Workspaces and Environments Workspaces are the top-level organizational unit in Agent Studio. This guide covers how to create a workspace and how to navigate between workspaces and environments in your day-to-day work. #### Before You Start * You need appropriate permissions to create workspaces in your tenant. * Plan your workspace structure before creating. You are limited to five workspaces per tenant. * Use clear, descriptive names and descriptions so other users understand the purpose of each workspace. ### How to Create a Workspace {% stepper %} {% step %} Log in to Agent Studio. {% endstep %} {% step %} Click the workspace name in the breadcrumb at the top left of any page, or click the Workspaces icon in the main navigation menu. {% endstep %} {% step %} In the workspace panel, select the option to create a new workspace. {% endstep %} {% step %} Enter a name for the workspace that clearly identifies the business area, site, or use case. {% endstep %} {% step %} Enter a description explaining what the workspace is intended for and who manages it. {% endstep %} {% step %} Save the workspace. It will appear in your workspace list immediately. {% endstep %} {% endstepper %} ### How to Switch Between Workspaces and Environments The breadcrumb at the top left always shows your current workspace and environment. * To switch workspaces: click the workspace name in the breadcrumb and select the desired workspace from the panel. * To switch environments within a workspace: click the environment name in the breadcrumb (UAT or Production) and confirm the switch. {% hint style="info" %} **Reminder:** Changes to Digital Workers, including installation, configuration, and publishing, always happen in UAT. Production is read-only except for HITL configuration, model parameters, and concurrency settings. {% endhint %} ### Key Rules | Rule | Detail | | -------------------------- | ----------------------------------------------------------------------------------------------- | | All installs happen in UAT | Digital Workers can only be installed into the UAT environment, never directly into Production. | | Changes stay in UAT | You can only edit the Draft version of a Digital Worker, which exists in UAT. | | Promotion is intentional | Moving a Digital Worker to Production requires a deliberate publish and promote action. | | Maximum 5 workspaces | Plan your workspace structure before you reach the limit. | ### Workspace Planning Tips * Create workspaces around business units, sites, or product areas that have distinct processes and data, not around individual Digital Workers or use cases within a single area. * Document the purpose and owner of each workspace in the description field so the team knows where new Digital Workers should be installed. * If you are close to the five-workspace limit, review whether any existing workspaces can be consolidated before creating a new one. # Customize Instructions Instructions are the core policy of a digital worker. Changing them is often the first and most impactful customization you can make. This guide covers how to update instructions effectively to shift the worker's behavior without requiring changes to skills or tools. **When to Change Instructions** * The digital worker's output format does not match what downstream systems or users expect. * The worker is not handling certain edge cases correctly. * You need to add, remove, or reorder steps in the worker's reasoning process. * You want the worker to ask different follow-up questions when information is missing. **Where to Find Instructions** 1. Open the digital worker from the Digital Workers page. 2. Navigate to the Compose tab. 3. Locate the instructions field for the digital worker or the specific skill you want to modify. You are working on the Draft version. Instructions on published versions are read-only. **How to Write Effective Instructions** Instructions should be specific, complete, and unambiguous. The digital worker follows instructions literally, so vague guidance produces inconsistent results. Effective instructions typically cover: * What inputs to accept and how to validate them. * How to reason through the task, including what to do when information is incomplete. * What format to return, including required fields, optional fields, and null-handling rules. * How to handle edge cases and exceptions. Example: If you want the worker to return structured JSON instead of a free-form summary, add an explicit output contract to the instructions. Specify the required keys, their data types, allowed values, and what the worker should set when a field is unknown (for example, null), along with a rule to ask exactly two follow-up questions when required fields are missing. **Using Meta Prompter for Instruction Changes** If you prefer not to edit instructions directly, the Meta Prompter provides a conversational alternative. Open the Meta Prompter from the digital worker's Graph view, describe the change you want to make, review the proposed plan, and confirm. The Meta Prompter applies the change to the Draft for you. **After Changing Instructions** 1. Run sanity tests from the Compose tab to verify the updated behavior. 2. Review outputs for accuracy, format, and edge case handling. 3. If results are as expected, publish the Draft to create a new version. 4. Promote the new version to Production when ready. *Best practice: Test instruction changes against a representative set of real or realistic inputs, including edge cases, before publishing. Small wording changes can have significant downstream effects.* # Set Up Scorecards **How to Set Up and Use Evaluation Scorecards** Evaluation scorecards let you define what good performance looks like for a digital worker and then measure every interaction against those criteria automatically. This guide walks through creating a scorecard, attaching rubrics, and reviewing results. **Step 1: Create a Scorecard** 1. Navigate to the evaluation section for the digital worker you want to measure. 2. Create a new scorecard and give it a descriptive name that reflects its purpose, for example Dispatch Triage Quality or Support Resolution Quality. 3. Add score categories. Categories are high-level measurement dimensions. Common examples include Policy and Process Adherence, Accuracy and Resolution Quality, Communication Quality, and Safety and Compliance. 4. Assign a weight to each category so the scores roll up into a meaningful overall score. For example: 40% accuracy, 30% process adherence, 20% communication, 10% compliance. Weights must total 100%. 5. Save and move on to creating rubric criteria for each category. **Step 2: Create Rubric Criteria** Rubrics are the specific checks used to score each interaction. For each category in your scorecard, create one or more rubric criteria following these principles: * Write each criterion as a clear, observable question or check, something a reviewer can answer consistently without ambiguity. * Choose a scoring scale for each criterion: Pass/Fail, 1 to 5, or 0 to 2. * Add reviewer guidance: describe what a good response looks like, list common failure modes, and note what evidence to look for in the conversation transcript. * Assign each criterion to its relevant scorecard category. **Step 3: Publish the Scorecard** 1. Review the scorecard structure, including categories, weights, and criteria, to confirm they reflect your quality goals. 2. Publish the scorecard to the relevant digital worker(s). 3. Start in UAT to validate that the scoring is working as intended before publishing to Production. **Step 4: Review Interaction-Level Results** After a conversation closes, scores are calculated automatically. To review results: 1. Open any agent interaction from the monitoring or conversation view. 2. Review the rubric panel on the right side of the screen. It shows criterion-level scores, category rollups, and the overall score. 3. Click into any failed criterion to see what evidence in the conversation drove the lower score. **Step 5: Monitor Trends in the Dashboard** 1. Open the Evals dashboard for the digital worker. 2. Review overall score trends over time (daily or weekly). 3. Use filters to slice by digital worker, workspace, environment, channel, score category, or individual rubric criterion. 4. Pay attention to distributions, specifically the percentage of interactions below threshold, not just average scores. 5. After releasing a new version, use the dashboard to validate whether the change improved quality. *Tip: Track rubric versions over time. If you change a criterion's definition or scoring scale, historical scores will no longer be directly comparable. Version your rubrics to maintain a reliable audit trail.* # HIDDEN - Resource library A complete collection of articles and videos covering Agent Studio in depth. Return to it as a reference whenever you need more detail on any topic.
Fundamentals * [**What Agentic Studio is & where it fits**](/agentstudio/resource-library/hidden-fundamentals/what-agentic-studio-is-and-where-it-fits) *Agentic Studio's role in the IFS Loops platform & how it connects to CX workflows, data sources, & external systems.* * [**How tenants & workspaces are structured**](/agentstudio/resource-library/hidden-fundamentals/how-tenants-and-workspaces-are-structured) *The two-level environment structure — tenants as the organizational boundary & workspaces as isolated areas within them.* * [**How to navigate the Agentic Studio interface**](/agentstudio/resource-library/hidden-fundamentals/how-to-navigate-the-agentic-studio-interface) *The main navigation areas, menu structure, & where to find key features across the interface.* * [**What Digital Workers are & how they work**](/agentstudio/resource-library/hidden-fundamentals/what-digital-workers-are-and-how-they-work) *What Digital Workers are, how they're structured, & how their components work together at runtime.*
Core concepts * [**What tools are & how Digital Workers use them**](/agentstudio/resource-library/hidden-core-concepts/what-tools-are-and-how-digital-workers-use-them) *What tools are, the types available, & how Digital Workers use them to retrieve data, run logic, & interact with external systems.* * [**What connectors are & how they work**](/agentstudio/resource-library/hidden-core-concepts/what-connectors-are-and-how-they-work) *How connectors link Agentic Studio to external platforms & how authentication & communication are handled.* * [**How memory works in Agentic Studio**](/agentstudio/resource-library/hidden-core-concepts/how-memory-works-in-agentic-studio) *The memory types available to Digital Workers & how each supports different retention needs across interactions.* * [**What skills are & how they extend Digital Workers**](/agentstudio/resource-library/hidden-core-concepts/what-skills-are-and-how-they-extend-digital-workers) *What skills are & how they work as reusable, shareable capabilities across Digital Workers in a workspace.* * [**How evaluations work & what they measure**](/agentstudio/resource-library/hidden-core-concepts/how-evaluations-work-and-what-they-measure) *How evaluations assess performance & what the scored results indicate about behavior.*
Managing Digital Workers * [**How to import a Digital Worker from the Marketplace**](/agentstudio/resource-library/hidden-managing-digital-workers/how-to-import-a-digital-worker-from-the-marketplace) *How to browse the Marketplace, select a Digital Worker, and add it to your workspace.* * [**How to configure a Digital Worker**](/agentstudio/resource-library/hidden-managing-digital-workers/how-to-configure-a-digital-worker) *How to open, modify, & save changes to a Digital Worker in your workspace.*
MetaPrompter * [**What MetaPrompter is & when to use it**](/agentstudio/resource-library/hidden-metaprompter/what-metaprompter-is-and-when-to-use-it) *MetaPrompter's role in Digital Worker development & the kinds of problems it's designed to solve.* * [**How MetaPrompter generates proposed changes**](/agentstudio/resource-library/hidden-metaprompter/how-metaprompter-generates-proposed-changes) *How MetaPrompter takes a behavioral description & turns it into targeted configuration proposals.* * [**The MetaPrompter interface: fields & controls**](/agentstudio/resource-library/hidden-metaprompter/the-metaprompter-interface-fields-and-controls) *The input area, output area, & proposal controls that make up the MetaPrompter interface.* * [**How to configure a Digital Worker with MetaPrompter**](/agentstudio/resource-library/hidden-metaprompter/how-to-configure-a-digital-worker-with-metaprompter) *How to use MetaPrompter to configure a Digital Worker.* * [**How to review & apply MetaPrompter changes**](/agentstudio/resource-library/hidden-metaprompter/how-to-review-and-apply-metaprompter-changes) *How proposed changes are displayed, compared against current configuration, & selectively applied.*
Graph view * [**What the graph view shows & how to read it**](/agentstudio/resource-library/hidden-graph-view/what-the-graph-view-shows-and-how-to-read-it) *Graph view as a visual map of Digital Worker structure — what nodes represent & how connections reflect runtime data flow.* * [**Node states in graph view & what they mean**](/agentstudio/resource-library/hidden-graph-view/node-states-in-graph-view-and-what-they-mean) *Each node state — idle, active, completed, & error — & what each indicates during Digital Worker execution.* * [**How to navigate nodes in graph view**](/agentstudio/resource-library/hidden-graph-view/how-to-navigate-nodes-in-graph-view) *Node selection, the detail panel, & how to move through complex Digital Worker structures using zoom & pan.*
# HIDDEN - Fundamentals * [**What Agentic Studio is & where it fits**](/agentstudio/resource-library/hidden-fundamentals/what-agentic-studio-is-and-where-it-fits) *Agentic Studio's role in the IFS Loops platform & how it connects to CX workflows, data sources, & external systems.* * [**How tenants & workspaces are structured**](/agentstudio/resource-library/hidden-fundamentals/how-tenants-and-workspaces-are-structured) *The two-level environment structure — tenants as the organizational boundary & workspaces as isolated areas within them.* * [**How to navigate the Agentic Studio interface**](/agentstudio/resource-library/hidden-fundamentals/how-to-navigate-the-agentic-studio-interface) *The main navigation areas, menu structure, & where to find key features across the interface.* * [**What Digital Workers are & how they work**](/agentstudio/resource-library/hidden-fundamentals/what-digital-workers-are-and-how-they-work) *What Digital Workers are, how they're structured, & how their components work together at runtime.* # What Agentic Studio is & where it fits ## Overview Agentic Studio is an all-in-one platform for building, deploying, and monitoring AI-powered Digital Workers at enterprise scale. It provides a structured environment for the full Digital Worker lifecycle, from initial configuration through testing, deployment, and continuous improvement, with end-to-end visibility and strong integration depth throughout. The platform is designed for both technical and non-technical users. AI and ML engineers, operations managers, compliance leads, business process owners, and platform administrators all work within Agentic Studio, each with different goals but a shared environment built around transparency, control, and fast iteration. ### What Agentic Studio Provides **A complete build environment.** Digital Workers are configured with instructions, tools, skills, triggers, and connectors, all within a single platform. Rather than assembling Digital Workers across separate systems, everything needed to define how a Digital Worker behaves, what it can access, and when it acts is managed in one place. **Connections to enterprise systems.** Agentic Studio connects to more than 70 enterprise applications, databases, data warehouses, and knowledge sources. These integrations give Digital Workers access to the data and systems they need to perform real work across enterprise workflows. **Structured testing and evaluation.** Before any Digital Worker reaches production, it passes through a dedicated evaluation environment where behavior can be validated interactively or against broader test datasets. Evaluation metrics, including accuracy, relevance, and helpfulness, are tracked and visible throughout. **Governed deployment.** Digital Workers move from test to production through a controlled release process. Publishing creates a versioned snapshot; pushing to production promotes that version into the live environment. Production workers are stable and not directly editable, protecting live behavior from unintended changes. **Continuous improvement.** Every interaction generates signals, including feedback, resolution outcomes, and recurring patterns, that feed back into improving Digital Worker performance over time. The platform supports structured learning through memory and fine-tuning capabilities that allow workers to become more effective as they accumulate operational experience. ### Core Design Principles Agentic Studio is built around four principles that shape how the platform works and how Digital Workers are developed within it. **Transparency first.** Every decision, tool call, and outcome is traceable and explainable. Users can inspect what a Digital Worker did, why, and with what result. **Composability**. Digital Workers are assembled from reusable building blocks, including instructions, skills, tools, triggers, and connectors. Components can be shared, reused, and updated across workers without rebuilding from scratch. **Safety and control.** Governance gates and human-in-the-loop patterns are built into the platform for actions that require oversight. Risky actions can be gated behind human review before execution. **Fast iteration.** The test, evaluate, learn, and release loop is a first-class part of the platform. Changes can be made in a safe environment, evaluated against structured criteria, and promoted to production when ready. # How tenants & workspaces are structured ### Overview Agentic Studio is organized into two levels of structure, tenants and workspaces, that together define how users, Digital Workers, and resources are separated and shared across an organization. Understanding this structure is important for anyone configuring, managing, or deploying Digital Workers at scale. ### Tenants A tenant is the top-level organizational unit in Agentic Studio. It represents a discrete deployment environment, typically corresponding to a company or a major organizational boundary, and contains everything associated with that environment: workspaces, users, connectors, and Digital Worker assets. Resources and data are isolated at the tenant level. One tenant cannot access the assets or configurations of another. Within a tenant, certain assets have visibility across all workspaces. Templates exported by users from any workspace within a tenant are visible to all other workspaces in that same tenant. This makes it possible to standardize and share Digital Worker patterns across teams without duplicating effort. ### Workspaces A workspace is a scoped environment within a tenant where Digital Workers are built, tested, and deployed. Each workspace provides an isolated context for development work, with its own Digital Workers, tools, skills, and configuration. Workspaces are where the day-to-day work of building and managing Digital Workers happens. Teams can operate within a workspace without affecting other workspaces in the same tenant, which supports parallel development and clean separation between different business functions, products, or use cases. ### Test and Production Environments Within each workspace, Agentic Studio maintains a clear separation between the test environment and the production environment. The test environment is the default working state. It is where Digital Workers are configured, evaluated, and refined before release. All changes to instructions, tools, skills, triggers, and settings are made in test. This keeps iteration separate from live operations and ensures that nothing reaches production without going through a deliberate release step. The production environment hosts deployed Digital Workers that are available for live use. Once a Digital Worker is promoted to production, it operates as a stable runtime artifact and is not directly editable. Any further changes are made in the test environment and promoted through the same controlled release process. This separation reduces the risk of accidental changes to live workers and provides a clear, auditable path from development to deployment. # How to navigate the Agentic Studio interface ### Overview Agentic Studio is organized around the Digital Worker lifecycle. The platform's main areas correspond to the stages a Digital Worker moves through, from initial creation and configuration, through testing and evaluation, to deployment and ongoing monitoring. This article describes what each area of the platform contains and when to use it. ### Digital Workers The Digital Workers area is the starting point for all development work. From here, users create new Digital Workers, import existing configurations from the marketplace or saved templates, and access the workspaces where individual workers are configured. Each Digital Worker has its own workspace, which contains the full set of configuration options: instructions, tools, skills, triggers, connectors, and settings. This is where a Digital Worker's behavior is defined and refined before it moves to testing. ### Compose The Compose area is where a Digital Worker's core configuration is built and edited. It contains the instructions that define how the worker behaves, the tools it can call, the skills it applies, and the triggers that initiate its activity. Settings such as reasoning effort, verbosity, human-in-the-loop behavior, and concurrency are also managed here. Changes made in Compose apply to the draft version of the Digital Worker. The draft is the active working state, editable, testable, and separate from the published production version. ### Evaluate The Evaluate area provides a dedicated environment for validating Digital Worker behavior before deployment. It supports two modes of testing: interactive testing for individual scenarios and ad hoc validation, and dataset-based testing for broader, repeatable evaluation across many cases. Interactive tests allow users to engage with a Digital Worker directly through a conversational interface, observe how it responds, and assess whether its behavior matches expectations. Evaluation signals, including accuracy, relevance, and helpfulness, are surfaced for each interaction. Dataset testing allows users to run a saved collection of test cases against the Digital Worker and review aggregate metrics across the full set. Datasets function as persistent evaluation assets that can be reused as the Digital Worker evolves, supporting regression testing and release readiness checks. ### Deploy The Deploy area manages the release of Digital Workers from test to production. Once a Digital Worker has been validated, the release process moves through two steps: publishing a new version and pushing that version to production. Publishing creates a versioned, release-ready snapshot of the current draft. Pushing to production promotes that snapshot into the live environment, where the Digital Worker becomes available for active use. The version history created through this process supports auditability and, where needed, rollback. ### Monitoring and Analytics The Monitoring and Analytics area provides visibility into how Digital Workers are performing in production. It covers operational metrics such as conversation volumes, completion rates, and user feedback, as well as quality metrics including compliance scores and conversation quality evaluations. Knowledge dashboards surface trending topics and common queries, giving teams insight into what users are asking and where patterns are emerging. Custom dashboards can be configured for teams with specific reporting needs. ### Connectors The Connectors area is where integrations with external systems are set up and managed. Connectors provide authenticated access to the enterprise applications, knowledge sources, databases, and APIs that Digital Workers rely on. A connector is configured once and can be reused across multiple tools and workers within the same workspace. # What Digital Workers are & how they work ### Overview A Digital Worker is Agentic Studio's core unit of AI capability. Each Digital Worker is a configured, goal-directed system that carries out specific business tasks, including retrieving information, taking action in connected systems, managing workflows, and interacting with users. Its behavior and scope are defined by the instructions and tools it is configured with. Digital Workers are not general-purpose assistants. They are purpose-built for specific domains and workflows, equipped with the capabilities they need to operate reliably in those contexts. ### What a Digital Worker Is A Digital Worker is defined by a combination of components that together determine what it knows, what it can do, and how it behaves. **Instructions** establish the worker's purpose, persona, constraints, and approach. They define what the worker is for, how it should respond, and what it should not do. Instructions are the foundational configuration layer; everything else operates within the boundaries they set. **Tools** give the worker the ability to act. A tool is a callable capability that allows the worker to retrieve information, execute operations, or interact with external systems. Tools extend what a Digital Worker can do beyond language reasoning alone, enabling it to take real action in connected enterprise environments. **Skills** extend the worker's reasoning. A skill is a reusable, domain-specific capability, such as a prompt template, a reasoning framework, or a structured task pattern, that improves the consistency and quality of the worker's outputs for complex tasks. Where tools handle action, skills handle cognition. **Triggers** define when the worker activates. A Digital Worker can be initiated by a user interaction, an incoming event, an API call, or a scheduled condition. Triggers connect the worker to the operational context in which it needs to act. **Connectors** provide the integration access that tools rely on. A connector establishes authenticated access to an external system, such as a CRM, a service management platform, a knowledge base, or a data store, and makes that system available for the worker to use.
# HIDDEN - Core concepts * [**What tools are & how Digital Workers use them**](/agentstudio/resource-library/hidden-core-concepts/what-tools-are-and-how-digital-workers-use-them) *What tools are, the types available, & how Digital Workers use them to retrieve data, run logic, & interact with external systems.* * [**What connectors are & how they work**](/agentstudio/resource-library/hidden-core-concepts/what-connectors-are-and-how-they-work) *How connectors link Agentic Studio to external platforms & how authentication & communication are handled.* * [**How memory works in Agentic Studio**](/agentstudio/resource-library/hidden-core-concepts/how-memory-works-in-agentic-studio) *The memory types available to Digital Workers & how each supports different retention needs across interactions.* * [**What skills are & how they extend Digital Workers**](/agentstudio/resource-library/hidden-core-concepts/what-skills-are-and-how-they-extend-digital-workers) *What skills are & how they work as reusable, shareable capabilities across Digital Workers in a workspace.* * [**How evaluations work & what they measure**](/agentstudio/resource-library/hidden-core-concepts/how-evaluations-work-and-what-they-measure) *How evaluations assess performance & what the scored results indicate about behavior.* # What tools are & how Digital Workers use them ### Overview Tools are the mechanism through which Digital Workers take action. A tool is a callable capability that enables a Digital Worker to retrieve information, execute operations, or interact with external systems. Without tools, a Digital Worker can reason and respond but cannot act. Tools are what allow it to do real work across enterprise systems and workflows. Tools are modular and configurable. They are attached to a Digital Worker during configuration, and each tool defines what the worker can do, what inputs it requires, and what it returns. Multiple tools can be attached to a single Digital Worker, allowing it to handle multi-step tasks that span different systems. ### Tool Types Agentic Studio provides two categories of tools: system tools and integration tools. **System tools** are platform-native capabilities that support the Digital Worker's internal reasoning, orchestration, and task execution. They help the worker think, decide, transform information, or navigate a workflow. Rather than reaching external systems, system tools operate within the platform to support how the worker processes and structures its work. One example is the Content Resolution tool, which helps a Digital Worker find relevant information from connected knowledge sources, match that content to the request at hand, and pass it into the next step in the workflow. Within this tool, users configure which knowledge sources the worker should search across, with options including internal enterprise knowledge, external sources, code repositories, or a specific connected knowledge base. **Integration tools** are action-oriented or retrieval-oriented capabilities connected to external business systems. They allow a Digital Worker to move beyond reasoning and interact with operational software, sending an email, creating a ticket in a service management platform, posting to a messaging channel, or pulling data from a connected API. Integration tools are designed for real-world execution, enabling the worker to take meaningful action in the systems an organization already uses. ### Tools and Connectors Integration tools work in conjunction with connectors. A connector establishes and manages authenticated access to an external system. An integration tool defines a specific action performed through that access. The distinction matters in practice: a connector to Zendesk provides access to that system, while an integration tool built on that connector defines a specific action, such as creating a ticket or retrieving a record. One connector can support multiple tools, each representing a different action available through the same system access. ### How a Digital Worker Uses Tools When a Digital Worker receives a request, its reasoning engine determines whether fulfilling that request requires calling one or more tools. It plans the task, selects the appropriate tool, structures the required inputs, executes the tool call, and integrates the result into its response or next step. This process can chain across multiple tools in sequence, with outputs from one step feeding into the next. Tools operate as deterministic, trusted capabilities that supplement the worker's language reasoning. Where reasoning handles interpretation and judgment, tools handle execution, ensuring that actions taken against external systems are accurate, consistent, and auditable. Every tool execution is logged. Inputs, outputs, the endpoint called, and the parameters used are all captured, providing a complete record of what the worker did and how it did it. ### Choosing Tools for a Digital Worker The right set of tools depends on what the Digital Worker needs to accomplish. A useful starting point is to consider what information the worker needs to retrieve, what actions it needs to perform, which internal systems are involved, and which external platforms it needs to reach. A Digital Worker managing inventory, for example, would need tools to retrieve warehouse data and product information, create records in an internal ticketing system, and send notifications when stock thresholds are met. Each capability maps to a specific tool configured for that purpose. Agentic Studio supports tool discovery through search, category browsing, and examples of tools already in use by other Digital Workers in the same workspace. This makes it easier to identify relevant tools quickly and understand how similar workers have been equipped. # What connectors are & how they work ### Overview Tools are the mechanism through which Digital Workers take action. A tool is a callable capability that enables a Digital Worker to retrieve information, execute operations, or interact with external systems. Without tools, a Digital Worker can reason and respond but cannot act. Tools are what allow it to do real work across enterprise systems and workflows. Tools are modular and configurable. They are attached to a Digital Worker during configuration, and each tool defines what the worker can do, what inputs it requires, and what it returns. Multiple tools can be attached to a single Digital Worker, allowing it to handle multi-step tasks that span different systems. ### Tool Types Agentic Studio provides two categories of tools: system tools and integration tools. System tools are platform-native capabilities that support the Digital Worker's internal reasoning, orchestration, and task execution. They help the worker think, decide, transform information, or navigate a workflow. Rather than reaching external systems, system tools operate within the platform to support how the worker processes and structures its work. One example is the Content Resolution tool, which helps a Digital Worker find relevant information from connected knowledge sources, match that content to the request at hand, and pass it into the next step in the workflow. Within this tool, users configure which knowledge sources the worker should search across, with options including internal enterprise knowledge, external sources, code repositories, or a specific connected knowledge base. Integration tools are action-oriented or retrieval-oriented capabilities connected to external business systems. They allow a Digital Worker to move beyond reasoning and interact with operational software, sending an email, creating a ticket in a service management platform, posting to a messaging channel, or pulling data from a connected API. Integration tools are designed for real-world execution, enabling the worker to take meaningful action in the systems an organization already uses. ### Tools and Connectors Integration tools work in conjunction with connectors. A connector establishes and manages authenticated access to an external system. An integration tool defines a specific action performed through that access. The distinction matters in practice: a connector to Zendesk provides access to that system, while an integration tool built on that connector defines a specific action, such as creating a ticket or retrieving a record. One connector can support multiple tools, each representing a different action available through the same system access. ### How a Digital Worker Uses Tools When a Digital Worker receives a request, its reasoning engine determines whether fulfilling that request requires calling one or more tools. It plans the task, selects the appropriate tool, structures the required inputs, executes the tool call, and integrates the result into its response or next step. This process can chain across multiple tools in sequence, with outputs from one step feeding into the next. Tools operate as deterministic, trusted capabilities that supplement the worker's language reasoning. Where reasoning handles interpretation and judgment, tools handle execution, ensuring that actions taken against external systems are accurate, consistent, and auditable. Every tool execution is logged. Inputs, outputs, the endpoint called, and the parameters used are all captured, providing a complete record of what the worker did and how it did it. ### Choosing Tools for a Digital Worker The right set of tools depends on what the Digital Worker needs to accomplish. A useful starting point is to consider what information the worker needs to retrieve, what actions it needs to perform, which internal systems are involved, and which external platforms it needs to reach. A Digital Worker managing inventory, for example, would need tools to retrieve warehouse data and product information, create records in an internal ticketing system, and send notifications when stock thresholds are met. Each capability maps to a specific tool configured for that purpose. Agentic Studio supports tool discovery through search, category browsing, and examples of tools already in use by other Digital Workers in the same workspace. This makes it easier to identify relevant tools quickly and understand how similar workers have been equipped.
# How memory works in Agentic Studio ### Overview Memory allows Digital Workers to retain and apply knowledge across interactions rather than treating each task as a fresh start. Without memory, a Digital Worker processes every request in isolation, with no awareness of prior context, past decisions, or accumulated operational experience. With memory, a worker can build continuity over time, applying what it has learned from previous interactions to improve the consistency and quality of its work. Agentic Studio organizes Digital Worker memory into a structured three-plane architecture that separates different types of retained knowledge so each can be stored, accessed, and applied appropriately. ### Memory Maps Memory in Agentic Studio is organized through Memory Maps. A Memory Map is the structured representation of everything a Digital Worker retains across interactions. Rather than storing all knowledge in a single undifferentiated store, Memory Maps separate retained knowledge into three distinct planes, episodic, procedural, and semantic, each designed to hold a different type of information and serve a different purpose during task execution. This structure is what makes retained knowledge usable rather than simply stored. When a Digital Worker draws on memory, it pulls from the appropriate plane depending on what the task requires: what happened before, how similar work was completed, or what the business context means. ### The Three Memory Planes #### Episodic Memory Episodic memory stores specific interactions and events. It allows a Digital Worker to remember what happened in prior sessions, who it interacted with, what decisions were made, what actions were taken, and what outcomes resulted.
This gives the worker continuity across time. When a case, asset, or user has a history of prior interactions, the worker can respond with awareness of that context rather than starting from zero. Prior escalations, earlier resolutions, and past conversations tied to the same record are all available through episodic memory. #### Procedural Memory Procedural memory stores learned procedures and workflows. It captures the patterns, sequences, and operational approaches that have proven effective for similar tasks. This allows the Digital Worker to retain the how of work, resolution flows, approval paths, escalation patterns, and recurring execution steps, and apply them more consistently over time. Rather than reasoning through a familiar process from scratch on each occasion, the worker can draw on established procedural knowledge to operate more efficiently and reliably. #### Semantic Memory Semantic memory stores general knowledge and concepts. It allows the Digital Worker to retain domain knowledge, business rules, entity relationships, and conceptual understanding of the environment it operates in. This includes information such as policies, operating thresholds, site hierarchies, vendor relationships, and business definitions. Semantic memory enables the worker to reason using durable, context-specific knowledge rather than relying only on what is present in the current session. ### How the Three Planes Work Together Each memory plane serves a distinct function, but they work together during task execution. Episodic memory tells the worker what has happened before. Procedural memory tells it how similar situations have been handled. Semantic memory tells it what the relevant business context means. A Digital Worker handling a recurring issue, for example, can recognize the case from prior interactions through episodic memory, apply the resolution approach that has worked before through procedural memory, and reason about the business rules and thresholds involved through semantic memory. The result is a worker that operates with increasing judgment and consistency as it accumulates experience. ### Memory and Continuous Improvement Memory is the foundation of how Digital Workers improve over time. Each interaction contributes to what the worker retains, and that retained knowledge shapes how it approaches future tasks. This is distinct from one-time configuration. Rather than being defined entirely at setup, a worker's effectiveness can grow as it encounters more situations, resolves more cases, and refines its understanding of the operational environment. # What skills are & how they extend Digital Workers ### Overview Skills are reusable, domain-specific capabilities that extend a Digital Worker's reasoning, structure, and output quality for complex tasks. Where tools give a Digital Worker the ability to act on external systems, skills improve how it thinks, providing structured patterns, reasoning frameworks, and domain knowledge that go beyond what a single instruction prompt can reliably achieve. A skill encodes repeatable operational intelligence. Rather than relying on the underlying language model to reason through a complex task from scratch each time, a Digital Worker equipped with the right skills can apply proven patterns consistently, producing more reliable and predictable results across workflows. ### How Skills Differ from Tools Tools and skills both extend what a Digital Worker can do, but they operate at different layers. A tool is a callable action that connects the worker to an external system, retrieves information, or executes an operation. A skill is a cognitive capability that shapes how the worker reasons, structures its thinking, and produces output. In practice, a Digital Worker often uses both in combination. A skill might guide how the worker analyzes a situation or formats a response, while a tool executes the resulting action in a connected system. Skills handle cognition; tools handle execution. ### Skill Types Agentic Studio supports four types of skills, each serving a different purpose in how a Digital Worker approaches its work. Prompt skills are reusable templates for language-based tasks such as summarization, classification, translation, rewriting, and report generation. They define the structure and format the worker should follow for a given type of output, producing consistent results across runs without requiring the full reasoning overhead of a more complex skill. Reasoning skills provide structured task breakdowns, planning heuristics, evaluators, and error correction patterns. They are applied to multi-step workflows where the worker needs to decompose a complex problem, plan a sequence of actions, or verify the quality of its own outputs before proceeding. Reasoning skills make behavior more reliable in scenarios where a single prompt is not sufficient to guide the worker through the full task. Domain skills are patterns tailored to specific fields such as logistics, finance, operations, supply chain, and field service. Rather than encoding general reasoning, they carry operational knowledge specific to a business context, reducing the amount of domain-specific instruction that needs to be embedded in the worker's core configuration and improving output quality in specialist environments. Composable skills are modular capabilities that other Digital Workers or tools can call directly. They enable skills to function as shared building blocks across a system of workers, so that a capability developed for one worker can be reused by others without duplication. This supports more scalable agent architectures where consistent, governed reasoning patterns are applied across multiple workflows. ### Why Skills Matter Without skills, complex tasks depend heavily on the language model's general reasoning ability, which can produce inconsistent or unpredictable results, particularly for tasks that require a specific structure, domain knowledge, or multi-step logic. Skills address this by encoding the repeatable patterns that produce reliable outcomes, turning general-purpose language reasoning into controlled, predictable behavior. Skills also improve maintainability. Because a skill is a shared, reusable asset, updating it once improves every Digital Worker that uses it. This is more efficient than updating instructions across multiple individual workers and reduces the risk of inconsistency between workers handling similar tasks. For organizations deploying Digital Workers across multiple teams or workflows, skills are the mechanism that ensures consistent reasoning standards are applied broadly rather than varying from worker to worker based on how each was individually configured. ### Skills and the Digital Worker Configuration Skills are attached to a Digital Worker during configuration alongside instructions, tools, triggers, and connectors. They operate as part of the worker's overall capability set, available to be invoked when the task at hand calls for the structured reasoning or domain knowledge they provide. A Digital Worker can be equipped with multiple skills, each addressing a different aspect of its work. The combination of skills a worker carries defines the depth and reliability of its reasoning across the range of tasks it is expected to handle. # How evaluations work & what they measure ### Overview The Evaluate section in Agentic Studio provides a dedicated environment for validating Digital Worker behavior before deployment and tracking performance over time. It is designed to support both fast, iterative testing of individual scenarios and structured, repeatable evaluation across larger sets of test cases. Evaluation in Agentic Studio helps answer questions that configuration alone cannot: whether the Digital Worker behaves correctly in a specific scenario, how it performs across a broader range of inputs, whether a recent change improved performance or introduced regressions, and which dimensions of quality need attention. The goal is to make testing observable, structured, and repeatable rather than dependent on informal spot checks. ### Two Modes of Evaluation Agentic Studio supports two evaluation modes that serve different purposes and work best at different stages of development. #### Interactive Testing Interactive testing is used for individual scenario validation, rapid iteration, and ad hoc exploration. It is the fastest path to understanding how a Digital Worker responds to a specific input, useful when validating a new configuration, testing the effect of a prompt change, exploring edge cases, or confirming that a particular workflow path behaves as expected. Within interactive testing, users can engage with the Digital Worker directly through a conversational interface, observing how it responds to real inputs in context. This simulates how the worker will behave in actual use and allows evaluation of tone, clarity, response structure, tool usage, and instruction adherence in a natural interaction flow. Interactive tests also serve as a source of reusable test cases. Individual interactions can be saved and added to a dataset, connecting exploratory testing directly to structured regression coverage. Agentic Studio supports two types of interactive tests. An event simulation test seeds the interaction with a real event payload to validate trigger behavior and the worker's response. A behavioral test evaluates agent logic, output format, or edge case handling independently of any specific trigger event. #### Dataset Testing Dataset testing is used for broader, repeatable evaluation across many test cases at once. A dataset is a saved collection of test cases that can be run against a Digital Worker and reused as the worker evolves. Rather than recreating the same scenarios each time a change is made, datasets provide a persistent evaluation asset that can be applied consistently across versions. Datasets are particularly useful for regression testing, release readiness checks, and performance tracking over time. Running a dataset after a configuration change makes it possible to see whether behavior has improved, degraded, or remained stable across the full range of cases, not just the ones that were manually tested in the current session. Datasets can be built from scratch or assembled by adding individual interactive test cases into a collection over time. This creates a natural workflow where exploratory testing generates the raw material for structured evaluation. ### Evaluation Metrics Both interactive tests and dataset runs surface evaluation metrics that make Digital Worker performance visible and comparable. The metrics available include accuracy, relevance, helpfulness, test success rate, failed tests, and average response time. In dataset testing, these metrics are aggregated across the full test suite, giving a summary view of performance at the collection level rather than requiring review of individual interactions. This makes it easier to assess overall readiness and identify patterns across a large number of cases. ### Out-of-Box and Custom Evaluations Agentic Studio provides out-of-box evaluations that are available for all Digital Workers without additional configuration, as well as the ability to define custom evaluations for specific use cases. Custom evaluations allow teams to assess performance against criteria that are particular to their workflows, outputs, or quality standards. An evaluation dashboard provides visibility into evaluation results over time, supporting ongoing quality monitoring beyond individual test runs. ### Evaluation in the Development Lifecycle Evaluation is not a single gate before deployment. It is a continuous part of the Digital Worker development lifecycle. Interactive testing supports rapid iteration during configuration. Dataset testing supports structured validation before release. Both feed into an ongoing loop where changes are tested, results are reviewed, and the worker is refined before any version reaches production. This approach keeps the path from development to deployment controlled and evidence-based, reducing the risk that changes which degrade performance go undetected before they affect live users. # HIDDEN - Managing Digital Workers * [**How to import a Digital Worker from the Marketplace**](/agentstudio/resource-library/hidden-managing-digital-workers/how-to-import-a-digital-worker-from-the-marketplace) *How to browse the Marketplace, select a Digital Worker, and add it to your workspace.* * [**How to configure a Digital Worker**](/agentstudio/resource-library/hidden-managing-digital-workers/how-to-configure-a-digital-worker) *How to open, modify, & save changes to a Digital Worker in your workspace.* --- [Next Page](/llms-full.txt/1)