AB-620: Designing and Building Integrated AI Solutions in Copilot Studio

By Microsoft · Question Mode

QuestionQ30

Test and manage agents

A Copilot Studio agent is assessed with a fixed test set and an automated evaluation method.

After the evaluation runs, the team finds that the same three interactions fail on every run, whereas all other interactions pass consistently.

You need to use the evaluation results to identify an accurate conclusion.

What should you conclude from the evaluation results?

Explanation

Consistent failures in particular test cases localize the problem to the corresponding portions of the agent’s functional scope, enabling targeted triage and remediation. Copilot Studio records pass/fail results for each test case and provides the response, analysis, and resources used, so isolated failures do not by themselves prove a language-model issue, require a complete redesign, or preclude deployment.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!