Agents and collaboration
Start and end a browser voice conversation
Use the Voice Testing screen to speak to an agent in the verified active project.
Before you begin
- Sign in and select the intended project.
- Live voice requires configured services/providers; demonstrations must use labelled deterministic boundaries.
Starting screen: /dashboard/agents
1. Open Voice Testing
In Agents open Configure Eva or an eligible specialist, then Voice Testing.
Expected result: The browser voice controls appear on an eligible plan.
Confirm it worked: Inspect the displayed state and verify the service-confirmed outcome before continuing.
2. Start a conversation
Choose Start browser conversation and grant microphone access if requested.
Expected result: The configured voice service establishes an authenticated session.
Confirm it worked: Inspect the displayed state and verify the service-confirmed outcome before continuing.
3. Follow transcript and tool activity
Speak a bounded business request. Read Live transcript and Voice tool activity where shown. Respond to questions/approval controls when required.
Expected result: Finalised transcript and tool outcomes are available; a transcript promise is not proof of a completed write.
Confirm it worked: Inspect the displayed state and verify the service-confirmed outcome before continuing.
4. End and inspect the call
Choose End conversation, then review Recent Calls after reconciliation.
Expected result: The session ends and its completed record can be inspected.
Confirm it worked: Inspect the displayed state and verify the service-confirmed outcome before continuing.
Troubleshooting and alternatives
Microphone access, a configured voice service and live provider availability are prerequisites. Illustrative demos use deterministic voice boundaries and must be labelled.
Owner voice writes follow policy and exact approval requirements. A declined or failed action must not be described as completed.