AI is in the thing you ship. This playbook takes an AI-powered product from problem to live to well-operated.

<aside> 📦

Full playbook on GitHub: step-by-step phases, exit criteria and templates. github.com/EmilieBuilds/AIProductPlaybook

</aside>

🔄 The lifecycle

Seven phases. If your team uses a standard product lifecycle, the middle column shows where you are.

Phase	In standard terms	The question it answers
1. Frame the problem 🔍	Define	Is this worth building, and does it need AI at all?
2. Govern before you build 🛡️	Plan	Who is accountable, and what are the rules?
3. Design the solution ✏️	Design	What is the simplest approach that could work?
4. Build 🔨	Build	How do we ship small and often?
5. Evaluate 🧪	Verify / Test	How do we know it works, and fails safely?
6. Launch 🚀	Release	How do we go live without betting everything on day one?
7. Operate and improve 🔁	Operate / Run	How do we keep it good after launch?

One difference from a standard lifecycle: this one loops. Evaluate can send you back to Design, and what you learn in Operate feeds the next round.

🧰 Templates

Fill-in documents included in the repo, used where they earn their place:

📄 Problem one-pager, Phase 1
⚠️ Risk register, Phase 2 onwards
🧪 Eval plan, Phases 3 to 5
✅ Launch checklist, Phase 6

🏁 How to use it

Read the overview. It takes 5 minutes.
Run Phase 1 on a real problem this week. Don't read the whole playbook first.
Skip what doesn't apply, but write down why you skipped it. That note is your audit trail.

<aside> ⚕️

Building in a regulated area? If your product gives clinical, financial or legal advice, or serves users in the EU, the ⚕️ steps are not optional. Get regulatory advice early. The playbook tells you when to ask, not what the answer is.

</aside>