1. Which four metric families describe a live AI system?
2. Which two metric families are most often absent, and whose job are they?
3. Why is "92% accurate" not a result?
4. What is the defining operational characteristic of AI systems?
5. Why can a fixed evaluation set fail to detect real degradation?
6. Which monitoring mechanism is described as the most valuable and most often skipped?
7. A live system's monthly cost has risen 30% over three months. Which figure determines whether this is a problem?
8. Why should a cost problem not be solved by restricting access?
9. What are the four conditions for scaling a system? Select all that apply.
10. What is the purpose of a post-implementation review?