Simulated Users and Multi Turn Conversations

How do you test an agent that cannot finish its task until a customer answers a question?

Some tickets reach the scheduling agent without the customer's windows. The agent must send a message asking when the customer is free, wait for the answer, and only then book. In the fake environment, nobody is there to answer, so the run stops at the question.

This lesson covers two ways to supply the missing customer, the errors that the pretend customer itself makes, and multi turn tests for features that are not agents at all.