Skip to main content
Testing is essential before deploying a Voice AI agent to live callers. This guide covers how to test your agent, interpret test results, and debug common issues.

​
Testing your agent

1

Use the built-in test call

In your Voice AI agent settings, click Test Call to initiate a test conversation. This lets you interact with your agent as a caller would, without using a live phone number.
2

Review the conversation transcript

After each test call, review the full transcript in the agent dashboard. Look for:
  • Correct understanding of caller intent
  • Appropriate responses and tone
  • Successful action execution (appointment booking, data collection)
  • Proper escalation behavior
3

Iterate on your prompt

Adjust your system prompt based on test results. Common improvements include:
  • Adding specific handling for edge cases
  • Clarifying the agent role and boundaries
  • Improving fallback responses
  • Refining appointment booking instructions

​
Common issues

Review your system prompt for clarity. Make sure the agent role and supported topics are explicitly defined. Add example scenarios for common caller requests.
Check your knowledge base and system prompt for outdated or conflicting information. Ensure business hours, pricing, and service details are current.
Verify your phone number is properly configured. Check that the speech-to-text and text-to-speech settings match your expected language and accent.
Confirm your calendar is connected and the agent has the correct calendar ID. Test that free slots are available for the time ranges your agent offers.

​
Next steps

Creating agents

Review agent configuration options

Call escalation

Set up human handoff for complex situations
Last modified on March 5, 2026