> For the complete documentation index, see [llms.txt](https://kb.theloops.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://kb.theloops.io/digitalworkers/ifs-loops-digital-workers/digital-worker-guides/closeout-and-improvement-manager/core-concepts/architecture-and-capabilities.md).

# Architecture & capabilities

## Overview&#x20;

The Closeout and Improvement Manager is built on five agentic skills that together take a completed maintenance work task from initial review through to a closed work order: Work Context & Quality Assessment, Work Validation & Authorization, Follow-up Work & Improvement Detection, Maintenance Decision Closeout, and Closeout Summary & Notification. This article covers those skills, the closeout workflow they share, integration points, and how the system keeps a supervisor in the loop on the decisions that need one.&#x20;

## Agentic Skills&#x20;

#### Work Context & Quality Assessment&#x20;

Detects completed work tasks and retrieves the work task, work order, asset, technician, and plan details, enriched with the asset’s history and past maintenance context. This skill also evaluates the technician’s remarks, actions taken, and findings, scoring documentation completeness and clarity so that quality is assessed the same way on every task.&#x20;

#### Work Validation & Authorization&#x20;

Compares reported labor, materials, services, tools, and equipment against what was planned, validating costs against configurable tolerance rules and identifying variances and exceptions. This skill also consolidates every validation result, applies business rules and approval policies, and auto-authorizes a work task once it passes all checks, routing anything else for human-in-the-loop review.&#x20;

#### Follow-up Work & Improvement Detection&#x20;

Analyzes notes and observations using AI to detect new defects, anomalies, or incomplete work, and to identify recurring issues or reliability risks. This skill recommends follow-up actions or improvement opportunities, including ones mentioned only in a technician’s own words rather than flagged through structured data.&#x20;

#### Maintenance Decision Closeout&#x20;

Creates a work request or fault report for a new issue, or a maintenance change request when the finding points to a process or plan improvement, and routes it for approval to the supervisor, planner, or reliability engineer. This skill also validates the parent work order: confirming every child task is completed and authorized, checking for pending follow-up actions or maintenance change requests, and validating the asset hierarchy and any serialized component updates before closeout.&#x20;

#### Closeout Summary & Notification&#x20;

Generates a closeout summary capturing key decisions and insights, updates the work order status to closed, records the audit trail and closeout history, and sends notifications through Microsoft Teams and email.&#x20;

#### Closeout Workflow&#x20;

The following table outlines the process from a completed work task through to a closed work order.&#x20;

| Stage                             | What the System Does                                                                         | Action / Output                               | Channel       |
| --------------------------------- | -------------------------------------------------------------------------------------------- | --------------------------------------------- | ------------- |
| Work Context & Validation         | Retrieves work task, order, asset, technician, and plan details, enriched with asset history | Work context assembled                        | IFS Cloud     |
| Documentation Quality Assessment  | Evaluates technician remarks, actions, and findings                                          | Documentation quality score determined        | IFS Cloud     |
| Planned vs Actual Validation      | Compares reported labor, materials, and costs against plan tolerances                        | Variances and exceptions identified           | IFS Cloud     |
| Follow-up & Improvement Detection | Analyzes notes with AI for defects, anomalies, and recurring issues                          | Follow-up actions or improvements recommended | IFS Cloud     |
| Maintenance Recommendation        | Creates a work request, fault report, or maintenance change request                          | Recommendation routed for approval            | IFS Cloud     |
| Work Task Authorization           | Consolidates validation results and applies approval policies                                | Task auto-authorized, or routed for review    | IFS Cloud     |
| Parent Work Order Validation      | Confirms child tasks and follow-ups are complete                                             | Work order confirmed ready for closeout       | IFS Cloud     |
| Work Order Closeout               | Generates closeout summary and updates work order status                                     | Work order closed, notifications sent         | Teams / Email |

## Integration Points&#x20;

**IFS Cloud (EAM/Maintenance):** The primary system of record. The Closeout and Improvement Manager reads work orders, work tasks, asset data, and maintenance history, and writes back authorizations, recommendations, and closeout status.&#x20;

**Microsoft Teams:** Used to send closeout notifications and route exceptions for human-in-the-loop review.&#x20;

**Outlook/Email:** Used alongside Teams for closeout notifications.&#x20;

**Other Enterprise Systems:** Additional integration points beyond core IFS Cloud EAM/Maintenance functionality, as configured for the deployment.&#x20;

## Human-in-the-Loop Design&#x20;

The Closeout and Improvement Manager is designed so that clean, fully validated work authorizes and closes automatically, while exceptions and genuine judgment calls surface to a supervisor with an explanation attached.&#x20;

The system authorizes and proceeds automatically when:&#x20;

* Reported labor, materials, services, and costs fall within the configured tolerance against plan&#x20;
* Documentation quality clears the required bar&#x20;
* No follow-up or improvement action is pending approval&#x20;

A supervisor, planner, or reliability engineer is brought in when:&#x20;

* A validation result falls outside tolerance and is routed as an exception&#x20;
* A follow-up work request, fault report, or maintenance change request needs approval&#x20;
* The parent work order has a pending follow-up action or maintenance change request still undecided&#x20;

Exceptions are surfaced with explanations, so a supervisor can approve, reject, or clarify a decision rather than having to reconstruct the reasoning themselves.&#x20;

## Feedback and Continuous Improvement&#x20;

Maintenance governance is enforced consistently: the same policies, approval rules, tolerance limits, and asset hierarchy validations apply to every work task, regardless of who completed it or how busy the supervisor reviewing it might be. Real-time operational visibility, through dashboards on work status, exception trends, improvements, and closeout performance, gives supervisors and reliability engineers insight into where closeout bottlenecks and recurring issues are showing up across the maintenance team, rather than that picture only emerging after the fact.&#x20;