When I dug deeper into the criteria Artificial QA uses to evaluate a conversational agent, I discovered there were 17 different criteria at play! Some are the ones you'd expect... but others are far more subtle than you'd think (and your chatbot probably isn't meeting them). From the most technical to the most human, including risk-related issues that very few people ever check. 
In today's video, I walk you through all 17 and explain why they matter so much for building a truly reliable conversational agent.
Frequently asked questions
How many criteria should you consider when evaluating a chatbot?
According to the criteria Artificial QA uses, there are 17: from the most technical to the most human, including risk-related topics that very few teams check.
Why isn't it enough to test that the chatbot answers correctly?
Because many of those criteria are subtler than they seem, and a truly reliable conversational agent has to meet them, not just give correct answers.
Keep reading: From generating test cases with AI to running them with Hootie Live Test and When the tester goes from executor to orchestrator. If you want to go deeper, I recommend the course Testing of artificial intelligence systems.