Troubleshooting common Voice AI issues
Learn how to identify and resolve common Voice AI data flow, connectivity, and configuration issues. When issues arise, start by identifying where in the process the failure occurs. Voice AI issues generally fall into two categories:
- Data flow issues: These issues occur early in the process and prevent the system from transcribing captured audio or sending data to the Pega server. If no transcription is visible on the agent desktop, investigate data flow issues first.
- Agent desktop issues: These issues occur when transcription is successful but the agent browser does not receive the expected suggestions, such as suggested cases, knowledge articles, or form autofill content. If transcripts are visible but suggestions are missing, the issue typically lies in the Voice AI channel configuration.
Diagnostic procedures
Use the following steps to systematically identify and resolve issues.
- Verify enablement.
Confirm that the Enable Voice AI checkbox is selected in the Interaction Portal under Administration > Voice AI.
- Test connectivity.
Run the VerifyVoiceAIDataRetrieval activity. A successful configuration returns an HTTP 202 or 204 response code. Any other result indicates a connectivity or configuration problem.
- Enable desktop logging.
Set the When Rule named EnableAIDesktopLogging to true. This setting logs browser events, utterances, and suggested actions to the browser console, making it easier to trace issues on the agent desktop.
- Enable server-side logging.
Use the EnableVoiceAIPartialUtteranceLogging and EnableVoiceAIFinalUtteranceLogging when rules to write transcript data and utterance counts to the PegaRules.log on the RealTime node. These logs help isolate whether the issue is in audio capture, transcription, or data delivery.
- Test topic configuration.
Use the Preview console within the Voice AI channel to test keywords and verify that specific topics are correctly detected before deployment.
Common issues and resolutions
The following table shows common issues with Voice AI operation.
| Issue | Likely cause | Resolution |
|---|---|---|
|
Suggestions not appearing |
Incorrect topic criteria or Pega GenAI™ Knowledge Buddy integration settings. |
Review topic configuration and verify Knowledge Buddy integration. |
|
Failed entity extraction |
Natural Language Processing (NLP) models are not properly trained, or business cases are not mapped to the conversational channel. |
Retrain NLP models and verify channel mapping. |
|
VoiceAIRetrieval data flow stopped |
Data flow failure. |
Restart the data flow by using the Voice AI System Health Dashboard. |
|
No transcription visible |
Voice AI not enabled or connectivity failure. |
Complete Steps 1 and 2 in the diagnostic procedures above. |
This Topic is available in the following Module:
Want to help us improve this content?