Failing tests because everything is recognised as our agent
We’re trying to test an answering agent through two text conversations, but the tests often fail because it seems the system is not discerning between who is the caller and who is the agent. It sees everything as the agent and so we end up failing every test, how could we solve this issue?
