Find the cause of a voice issue¶
Make a test call that reproduces the problem. Then check whether the Agent heard the caller correctly, found the right answer and completed the intended action.
Follow the call path¶
| Question | Where to check | Guide |
|---|---|---|
| Did the number reach the intended workflow? | Default route, override, timezone | Number routing |
| Did recognition capture the caller correctly? | Finalized transcript | Voice configuration |
| Did the Flow choose the right specialist? | Routing decision | Agentic Flows |
| Was the needed information available? | Resolved prompt, lookup and knowledge events | Debug Monitor |
| Did the action run with correct inputs? | Tool invocation and external result | Tool guidance |
| Did the final result arrive? | Stored summary and recipient system | Post-processing |
Keep the test controlled¶
Use the same number, caller statements and expected outcome before and after a change. Record which Agent or Flow and configuration were used. Start a fresh call after changing assignments or available tools.
Open the test call early in the Debug Monitor so you can follow the available events.
Report what failed¶
Describe what you expected and what happened instead. Include the call identifier and approximate time. If you use Copy context, remove credentials and unnecessary personal details before sharing it. A short excerpt is usually enough.
For a forwarding issue, include the attempt’s final status. For a missing email, also check the reported sending result.