AI-103 Certification Practice Question #38

Question

You have a Microsoft Foundry project that contains an agent. The agent generates summaries from retrieved policy documents.
You need to improve response completeness. The solution must be implemented in the logic of the application code before responses are returned.
What should you do?

Accepted Answer

The fix must live in the application code, before the response is returned, and must improve completeness. Evaluating each generated response (e.g., with the response_completeness evaluator) and retrying generation when it falls short is an in-code quality gate that directly targets completeness — the agent re-generates until the response is complete enough, then returns it. Decreasing `max_tokens` (B) shortens output and makes responses *less* complete; switching to RAG (C) is an architectural change to grounding rather than a completeness check in app logic; replacing with a smaller model (D) typically lowers answer quality. The Suggested Answer is A and the distractors clearly work against completeness, but confidence is held at 78% because "retry evaluation" is non-standard phrasing with no community votes to corroborate, leaving some ambiguity about the exact mechanism intended.

More AI-103 practice questions