Services / AI Delivery Setup
AI product operating model (spec-driven delivery)
Four to six weeks to set up a spec-driven, AI-led way of building product in your own tools, with one real feature shipped through it.
- Price
- From £15,000
- fixed price, four to six weeks
- Time
- 4–6 weeks
- For
- Product and engineering teams that want to ship more with the team they have
The spec is the source of truth. Strategy feeds the roadmap, the roadmap feeds structured specs, and from the specs AI generates tickets, acceptance criteria, test cases and prototypes. Humans own judgement, trade-offs and review. That’s the whole idea, and it gets more shipped per week without quality or alignment slipping.
Signs you need this
- Your engineers use coding agents, but the specs they're given aren't good enough for a human, let alone a machine.
- AI has made building faster and deciding slower. The bottleneck moved and nobody noticed.
- You've tried "AI for product management" and ended up with longer documents, not better decisions.
- You want the speed without losing control of what ships.
Not for you if
- You want AI to make the product decisions. It shouldn't, and this is built so that it can't.
- Nobody on the team will own the spec once I leave. We'll find that person first, or not start.
What you get
- A current-state picture of how ideas become shipped work today, and where the time goes.
- A spec standard: strategy, roadmap item, spec and ticket, each traceable to the one above.
- The toolchain set up in your tools (Claude or ChatGPT, Linear or Jira, Notion or Confluence, GitHub) with reusable prompts for writing, critiquing and ticketing specs.
- One real feature taken from idea to shipped through the new flow, with before-and-after numbers.
- Human review gates and a clear line on what AI must never decide.
- Training for the team and a playbook you keep.
How it works
- Weeks 1–2: current state, spec standard, toolchain set up.
- Weeks 3–5: the pilot feature goes through the flow with the team; we fix what breaks.
- Week 6: training, playbook, measurements, handover.
Questions people ask
- Which tools does this work with?
- Whatever you already use. The method is the point; the tools are interchangeable. I've built it on Claude, Linear, Notion and GitHub, and it ports.
- Does this replace product managers?
- No. It removes the mechanical work (tickets, acceptance criteria, test cases, first drafts) so the judgement work gets done properly.
- Have you done this for real?
- Yes. I built this pipeline inside a new venture for a PE-owned software group, taking the team from an ideas list to a working product with paying customers, and wrote a course on it called "Product Management in the Age of AI".
- Can we start with a workshop?
- Yes. The half-day workshop covers the method and is credited against the full setup.