# Understanding the Agent Development Life Cycle (ADLC)

The Agent Development Life Cycle, or ADLC, is the structured process Agentic Studio uses to manage how digital workers are built, validated, deployed, and improved over time. It is modelled 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** 

Agentic 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" %}
&#x20;*<mark style="color:$primary;">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.</mark>*&#x20;
{% endhint %}

### **What Can Be Changed in Production**&#x20;

Once a digital worker is running in Production, it is not editable. Three settings can be adjusted without republishing:&#x20;

* HITL configuration: update escalation rules or routing channels without creating a new version.&#x20;
* Model parameters: tune the underlying model settings for performance.&#x20;
* Parallelism and concurrency: adjust how many tasks the worker can process simultaneously.&#x20;

For any other change, return to UAT, update the Draft, publish a new version, and promote it to Production.&#x20;


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://kb.theloops.io/agenticstudio/resource-library/explainers/understanding-the-agent-development-life-cycle-adlc.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.