​
Evaluate agent performance with simulated sessions

Test your AI agent’s performance using automated simulated conversations to get insights into how well your agent handles different scenarios and user interactions.
This tests a refund scenario where:
  • Customer needs refund for defective product
  • Agent verifies purchase
  • Policy guides the process
  • Must resolve in 5 turns
1

Create a Dataset for testing

  • Configure the agent dataset template with:
  • Agent scenarios: Define specific situations for testing (e.g., “Update address”, “Order an iPhone”)
  • Expected steps: List expected actions and responses Agent Dataset template Agent Dataset sample data
2

Set up the Test Run

  • Navigate to your http endpoint, click “Test”, and select “Simulated session” mode
  • Pick your agent dataset from the dropdown
  • Configure additional parameters like persona, tools, and context sources
  • Enable relevant evaluators Configure simulation Test Run
3

Execute Test Run

  • Click “Trigger test run” to begin
  • The system simulates conversations for each scenario
4

Review results

  • Each session runs end-to-end for thorough evaluation
  • You’ll see detailed results for every scenario Simulation Test Run result