1. What is the central economic difference between an AI product and conventional software?
2. What should be included in cost to serve that teams routinely omit? Select all that apply.
3. Which cost lever is usually the largest single saving available?
4. Why is an unprofitable power user more serious in an AI product than in conventional software?
5. What are the five rungs of the capability ladder, in order?
6. What does tool use and orchestration buy that the lower rungs do not?
7. What test should be passed before moving up a rung?
8. What does fine-tuning cost you beyond the training effort itself? Select all that apply.
9. How should a fair-use bound be set and communicated?
10. What should generally be done when inference costs fall?