We are announcing the ability to evaluate voice agents by assessing spoken interactions through their text representations in Microsoft Copilot Studio. This feature will reach general availability on May 31, 2026.
How does this affect me?
This feature helps you:
- Evaluate voice-driven agent experiences using the same framework you use for chat.
- Apply consistent quality measurement across agent modalities (voice and text).
- Reduce effort when validating voice agents by evaluating transcribed interactions instead of handling audio directly.
With voice interactions automatically represented as text, you can score and compare sessions the same way you do for other agents, measuring quality, accuracy, and expected behavior without adding voice-specific complexity to your evaluation process.
What action do I need to take?
This message is for awareness, and no action is required.