Initech's support bot quotes wrong plan details, ignores the house reply style, and sends tickets to the wrong squad. Your call: a better prompt, retrieval (RAG) or fine-tuning, and in what order?
Key idea: A better prompt fixes rules you can write down, RAG fixes facts the model can't know, and fine-tuning teaches patterns from examples but bakes in old facts. Most real products combine them.
Finish a run to see what a PM takes away from this lesson.
Which approach fixes which problem
Run an approach to add it here
Things to try
- Run Plain. Which kinds of job does it get wrong?
- Run Better prompt. What did instructions fix, and what couldn't they fix?
- Run RAG. Why do the facts come right while nothing else changes?
- Run Fine-tuned. It knows the squads, but look at its prices. Where did they come from?
- Run All three. If you could ship only one approach, which would it be, and what would you give up?