Microsoft CopilotfeatureCopilot Studio Agent Evaluations (Generally Available)
Test a Copilot Studio agent against a set of questions and reference answers before you ship it.
Details
- You build test sets by uploading them, writing them by hand, or having them generated, then score results against reference answers
- Multi-turn tests measure how the agent performs across a realistic back-and-forth rather than one question at a time
- You can compare two versions of an agent side by side to spot where quality slipped
Best for
Checking an agent still behaves after you change its instructions
When to use it
You are about to publish a change and want proof it broke nothing
Who benefits
Anyone maintaining an agent their team depends on
Join the newsletter to get ahead with AI.
I share strategies, new modules, free guides, and tools to get you better at using AI in your work.
Leading a team? See team programs →