Your agent has 99 problems.
Hallucination is just 1.

Asking an AI agent a list of questions and reading the answers tells you how it behaves in a demo. Real users push harder. They change their minds, give it wrong information, try to trick it, and arrive in their hundreds at the same moment. In this paper we have highlighted 50 ways agents fail, and how to test for it before a customer finds it.

What’s inside
In this paper we pick out 50 problems your agents face and show how to catch them with simulation-based testing.