What it tests
Ring 1 answers: Can the agent complete its task correctly under expected inputs? This is happy-path testing. Clean inputs, clear intent, straightforward flows. If your agent can’t handle the ideal case, nothing else matters.Prerequisites
- Call flow must be configured
How scenarios are generated
Ring 1 scenarios are generated per path in your call flow. For each path (e.g., “user places order successfully”, “user asks for store hours”), the AI generates test cases with:- User perspective: A realistic second-person description of what the caller does
- Expected outcome: Step-by-step expected agent behavior along this path
- Intent: What the caller wants, from the agent’s perspective
What it catches
- Agent doesn’t follow the defined call flow
- Missing steps in the conversation
- Incorrect transitions between nodes
- Agent fails to reach the expected end state
- Information is collected in the wrong order
Example
For a pizza ordering agent on the “successful order” path:User perspective: You call to order a large pepperoni pizza for delivery. You provide your address when asked and pay with a credit card. Expected outcome: Agent greets → asks pickup/delivery → collects address → takes pizza order → confirms order → processes payment → provides delivery estimate.