Create and refine a scenario
Define the interaction you want the Test Agent and AUT to have.
A scenario gives a generated test its purpose: who the TA is, how the conversation should unfold, and what quality signals matter.
Configure the scenario
Open Scenario Configuration and set the core information:
- Choose the TA Role and Vertical.
- Optionally choose a Registered AUT. Selecting one sets and locks the TA to the opposite role and uses the AUT’s vertical.
- Describe the Call Driver — why the TA is calling.
- Choose the relevant Rubrics & Metrics.
A registered AUT is optional unless you select Policy Enforcement. Policy Enforcement requires an AUT with successfully extracted, ready policy rules before scenario generation can continue. On Agents, follow the displayed policy status. Existing behavior definitions are read-only; if the definition is missing or needs different rules, delete and recreate the registration with a corrected definition. Then extract the policies and regenerate this scenario before running it.
Generate and refine
Select Generate Scenario to create a starting draft. Review it for realistic language, clear goals, and the right constraints. Use Refine with Feedback or Regenerate with Feedback to improve the draft when it needs a different emphasis.
Save a useful scenario with Save to Library so it can be reused later.
Hand the scenario to the Runner
When it is ready, select Run Test with this Scenario. This opens the Runner with the scenario ready to use.
The scenario guides the interaction; the Runner chooses the test mode, direction, and final submission settings.