Question

A solutions architect is finalizing the test plan for a Copilot Studio agent that answers policy questions grounded on a SharePoint knowledge source. Business stakeholders want assurance that the agent's prompts follow Microsoft's recommended prompt design practices before go-live. During validation, the QA team notices that for certain edge-case questions the agent produces confident-sounding answers that are not supported by the grounding documents. Which validation approach most directly confirms whether the prompts adhere to Copilot prompt best practices for this issue?

Accepted Answer

Run a groundedness-focused test set that checks whether each response cites and stays within the provided knowledge, verifying the prompt instructs the model to answer only from grounded sources and to state when information is unavailable

Answer

Increase the agent's response token limit so it can include more context in each answer and reduce truncation errors

Answer

Measure average response latency across 1,000 concurrent test users to confirm the agent meets performance SLAs

Answer

Switch the agent from generative orchestration to classic topics so all responses come from authored dialogs only