Test and manage agents

Official references for this objective

• Microsoft — Run and validate an AI model migration for Copilot Studio agents - Microsoft Copilot Studio | Microsoft Learn

  Do this while the current model is still in place, because after the model changes the baseline can't be reconstructed.

• Microsoft — Create a single response test set - Microsoft Copilot Studio | Microsoft Learn

  All test methods, except general quality , require expected responses or keywords.

• Microsoft — Automate evaluations with the Power Platform API - Microsoft Copilot Studio | Microsoft Learn

  Find and retrieve the test set ID of the test set you want to use. Run the evaluation . Retrieve the results by using the

• Microsoft — Analyze autonomous agents - Microsoft Copilot Studio | Microsoft Learn

  select All to display metrics for both the main agent and child agents, Main agent for metrics about the main agent only, or Child agent

• Microsoft — Change the details of a test set - Microsoft Copilot Studio | Microsoft Learn

  Tool use Whether the test case used all or any the expected resources Single response Pass/fail Expected capabilities

• Microsoft — Decide when to upgrade an AI model for Copilot Studio agents - Microsoft Copilot Studio | Microsoft Learn

  Go to the agent's Settings page and turn on Continue using retired models in the Model section when the additional compatibility period is required.

• Microsoft — Microsoft Foundry evaluation | Microsoft Learn

  Pin datasets, model deployments, evaluator versions, and rubric versions when results must be comparable across runs.

• Microsoft — Trigger agent evaluations with connectors - Microsoft Copilot Studio | Microsoft Learn

  Evaluate Agent : Run an evaluation for an agent by using a specific test set Get Agent Test Run Details : Get status and results