Agent Evaluations
Evaluations help you test your AI agent’s responses and ensure it’s providing accurate, helpful information to candidates.This documentation is coming soon. Check back for detailed guides on running agent evaluations.
Why evaluate
- Verify your agent handles common questions correctly
- Test edge cases and tricky scenarios
- Measure improvement over time
- Catch issues before candidates do
Running evaluations
- Navigate to Agents > Evaluations in Studio
- Create test cases with expected behaviors
- Run the evaluation suite
- Review results and adjust guidance as needed
Evaluation types
| Type | Description |
|---|---|
| Accuracy | Does the agent provide correct information? |
| Helpfulness | Does the agent address the candidate’s need? |
| Safety | Does the agent avoid inappropriate responses? |
| Policy compliance | Does the agent follow your defined policies? |