How the Problem Gets Solved

This part has three chapters, and together they answer one question: from the moment a customer says "we want AI" to the moment a system is actually running in production, what does that path look like?

The three chapters form a single pipeline: each one's output is the next one's input. First, find the right problem (Chapter 6 — discovery isn't brainstorming, it's filtering: take the customer's vague wish list, widen it out, then run it through five filters until a single worthwhile entry point remains; pick the wrong problem, and the harder the team works afterward, the more completely that effort goes to waste). With the problem settled, build the first version that's actually "real" (Chapter 7 — a Minimum Viable Deployment, or MVD: real data, real users, a real baseline — measuring the actual gap is what this stage delivers). With the gap measured, you can finally answer the question the customer is bound to ask (Chapter 8 — "so when can this go live" isn't a technical moment, it's a two-way acceptance contract). Without a problem that survives Chapter 6's five filters, an MVD has nothing to be "minimal" about; without the baseline Chapter 7 measures, there's nothing to write on Chapter 8's acceptance sheet.

This part's first readers are people about to walk onto the customer site: vendor-side delivery engineers, the technical lead running the deployment, and the customer-side owner who has to ask questions — and answer them — at the review table. The engineering details of moving a system into production — data, permissions, evaluation, billing — belong to Part Three. How this business gets paid belongs to Part Five. This part covers exactly one stretch of road: from a vague wish to a signed acceptance.