Five phases. Five gates. A gate never passes with open items.
Each phase ends in a pass-or-fail gate with written criteria. Your sponsor signs each one, and each signature is a fee milestone. Here is the whole method.
Diagnose
Weeks 1-2
What happens
Interviews with your team and customers, won and lost deal analysis, a stack audit and an AI search baseline.
What you give
System access, interview time, data exports.
Gate 1 · Build plan
Your ICP, first offer and three starting plays are agreed, and the build plan with gate dates is signed.
Prescribe
Weeks 2-4
What happens
Positioning, the knowledge base Claude reads, the architecture, and the first plays designed on paper.
What you give
Founder time for reviews, documents and proof.
Gate 2 · Foundation
Claude answers 20 of 20 new-hire questions correctly from your knowledge base, and sample drafts score 4 of 5 or better with no factual errors.
Deploy
Weeks 4-9
What happens
Every tool connected to Claude, the plays built and tested, the website rebuilt, and the demand engine set up.
What you give
Admin rights and the tool budget.
Gate 3 · Build complete
Every tool passes its connection test, three plays work end to end in test, and the new website is live.
Instrument
Weeks 9-11
What happens
Dashboard and weekly report, rep training, a shadow week, then a live week, then widening to more plays.
What you give
Rep time for approvals.
Gate 4 · Engine live
At least 9 in 10 signals are handled on time, nothing is sent without approval, and the dashboard shows real data.
Transfer
Weeks 11-12
What happens
Handover pack, owner training, and access moved fully to your team.
What you give
A named engine owner.
Gate 5 · Handover
Your owner runs the weekly review alone and adds a new play without help.
Ten principles, every client
- Knowledge before automation. Nothing is automated until the knowledge base can answer what a new rep would ask.
- Claude is the hub, not one more tool. Every system connects to Claude directly, or it gets replaced.
- One job per tool, one system of record. Your CRM owns the record, whether that's HubSpot, Salesforce or Dynamics 365. Everything else writes back to it.
- States, not stages. The engine acts on what is true about an account now, and the signal that changed it.
- Signal to action. Every signal has a play, an owner and a response time. A signal with no play is noise.
- Agents draft, people decide. Anything that reaches a buyer is approved by a person first.
- Build in your accounts. Every tool, file, skill and agent lives in accounts you own.
- Instrument from day one. Every play is logged, so we can show what it produced.
- Tune every week. A 30-minute review decides what to keep, change or switch off.
- Start narrow, then widen. One ICP, one offer, three plays. Add more only when those convert.
Shadow first. Then live. Then wider.
Nothing goes straight to your buyers. The first week, Claude drafts while your team works as usual and compares. Every gap gets fixed in the knowledge base, not patched by hand.
- Week 1, shadow: plays draft only; reps compare and flag gaps daily
- Week 2, live: approved drafts get sent; response times are tracked
- Week 3, wider: the next signals, the social queue and outbound sequences switch on
- Week 4, first monthly review: keep, change or switch off each play
- Every play starts at level 1 (draft only) and only moves up when it has earned it
See where your engine would start
No account team and no hand-off to a junior analyst. A 30-minute review with the consultant who builds your engine. Not ready yet? Get your score first.