How often is your AI simply wrong, and how would you know? Five episodes that take you from counting made-up answers to one quality score you can defend in a meeting.

1. Hallucination rate · 35 min
2. Groundedness rate · 35 min
3. Verified answer accuracy · 35 min
4. Evaluation pass rate · 35 min
5. Weighted quality score · 35 min