Overview

Before enabling Autopilot, test your AI Customer Agent to make sure its responses are accurate, consistent with your configuration, and aligned with its objective.

Start with Try Me, then use Copilot for controlled live conversations before gradually moving to Autopilot.

Test with Try Me

Go to Agents, select your Agent, then open Basic Info > Try Me.

Try Me lets you role-play as a customer and test your Agent before relying on it for live automated conversations.

Ask several questions and continue the conversation for multiple messages instead of testing only one response.

Review each response

Check whether the Agent provides accurate information from its Knowledge and follows the communication style and instructions you’ve configured.

Pay attention to its tone, response length, objective, playbook, and any special instructions.

If the Agent gives an incorrect answer or behaves unexpectedly, update its Basic Info, Knowledge, or Strategies and test the same scenario again.

Test multiple conversation turns

Don’t evaluate your Agent using only individual questions.

Continue each test for several messages to see whether the Agent maintains context and continues working toward its objective.

Include follow-up questions, objections, unclear requests, and changes in customer intent.

Check for Knowledge gaps

Review your test results and real conversations for questions your Agent couldn’t answer accurately.

Add missing information to Knowledge, remove outdated or conflicting content, and create Strategies for recurring situations that require specific handling.

You can also ask DealOnca to help review conversations and identify potential Knowledge gaps.

Move to Autopilot gradually

Once your Agent performs consistently in Try Me and requires minimal correction in Copilot, you can begin using Autopilot for appropriate conversations.

Continue monitoring Inbox and Support Needed conversations after enabling Autopilot.

If you notice recurring problems, switch affected conversations back to Copilot, make the necessary changes, and test again before resuming automation.

Keep testing as your Agent changes

Testing isn’t only for your first launch.

Retest your Agent after important changes to its objective, Knowledge, Strategies, playbook, or other instructions. This helps ensure that new information and configuration changes produce the behavior you expect.