We are announcing the ability to evaluate the entirety of multi-turn conversations in Microsoft Copilot Studio. This feature will reach general availability on June 30, 2026.
How does this affect me?
This feature enables assessment of agent behavior across an entire dialogue rather than grading or evaluating isolated responses. Instead of evaluating single prompt-response pairs, the system analyzes the full conversational flow.
This feature provides the following benefits:
- Improves evaluation accuracy by validating agent quality across full conversational flows, not isolated responses.
- Reduces production risk by detecting context loss, instruction drift, and breakdowns that only appear over multiple turns.
- Enables more realistic testing that mirrors real customer interactions..</li
- Accelerates issue identification in complex workflows, reducing costly post-release fixes..
- Strengthens release confidence for enterprise agents operating in multi-step scenarios..
This message is for awareness, and no action is required.
If you would like more information on this feature, please visit Evaluate the entirety of multi-turn conversations.
Source: Microsoft
Latest Posts
- MC1477184: Microsoft 365 Admin Center Consolidates Duplicate Agents and Adds Version History in Agent Registry
- MC1477185: OneDrive Adds Pay-As-You-Go Billing for Extra Storage
- MC1477186: Agent 365 Simplifies CLI Setup with Microsoft-Managed Enterprise Application
- MC1477187: OneDrive on macOS Increases Sync Limit to 1 Million Items per Device
