Train and Test helps admins check whether the AI Agent can recognize when it should act before deployment.
Before you start
Prepare a few recent tickets that represent the AI Agent's intended use. Include examples where the AI Agent should act and examples where it should not act.
Test the AI Agent
- Complete the AI Agent's Onboard step.
- Open the training or testing step when it is available.
- Use representative ticket examples for the selected categories.
- Review the AI Agent's suggested output or decision.
- Adjust categories, setup fields, or custom instructions if the result is wrong or incomplete.
- Repeat testing until the output matches the expected behavior.
What to review
- The AI Agent identifies the right tickets.
- The AI Agent ignores tickets outside the selected categories and configured conditions.
- Tags, statuses, priorities, custom fields, notes, or created tickets match the expected result.
- Edge cases and exceptions are covered in custom instructions.
What happens next
The AI Agent is ready for deployment with a clearer scope and fewer surprises during early activity.
Deploy in Supervised Mode when available so support agents can review early results.