Skip to content
methodology

how I think about systems.

Every partnership follows a process. I extract what you know, map how your business actually runs, and turn it into something that works without you having to explain it again.

C
R
A
F
T
cycle
the playbook90% of the result

your instructions. your methodology. your voice. your standards.

The instruction set that makes AI produce output worth using. You own it forever.

the tech10% of the result

tools change. platforms change. pricing changes.

Infrastructure you rent. A strong playbook works in any tool.

how it maps
processe.g. client onboarding

A core area of your business

workflowe.g. new client welcome sequence

A recurring sequence of steps inside a process

taske.g. send personalized welcome email

A specific action inside the workflow

playbooke.g. welcome email playbook

The AI instruction set for that task

01 · the build process

every process ships through the CRAFT cycle.

5 stages. Stage gates, not fixed timelines. Nothing ships until you test it on real work. One cycle per process, 2 to 4 weeks.

Each stage starts with the 5Ws: who, what, where, when, why.

I interview you for 60 minutes. Not about AI. About how you actually work. Your methodology, the exact words you use with clients, the tasks eating your time, what good looks like to you. By the end of the day, I have a draft of your AI configuration file. It reads like a manual only you could have written.

week 1

I check what is actually feasible before building anything. Which tasks are worth automating versus which ones genuinely need you. What the first version should do. Where the system hands work back to you. High value and simple to build goes first. Everything else waits. You sign off before I build anything.

week 1

I build the system. Your configuration file gets finalized, task instructions get written and tested, AI playbooks get wired together. Everything tested with real scenarios, not demos. Nothing moves forward until it works the way it should.

week 2

You use the system on real work. Every LinkedIn post, every client prep, every email. Feedback gets triaged into major and minor issues. Minor fixes are resolved immediately. Major issues get scoped into the next wave. This is the stage most people skip. It is the stage that makes the difference.

week 3

I walk through every trigger and output with you. You receive a step-by-step guide and a walkthrough video. Then a training call where you run every command yourself. You could hand this to a new team member and they would know exactly how to operate it.

week 3
waves

Each process gets its own CRAFT cycle. We call these waves. Each one builds on the last.

1
full build
2
faster
3
compounds
02 · how the AI thinks

four patterns. every workflow uses one.

The pattern determines whether AI produces output worth using.

guided generator

Turns a blank page into a finished draft

Follows your voice rules and structure to produce output. You edit, approve, publish.

LinkedIn posts · proposals · emails · newsletters

rubric-based evaluator

Scores against your criteria

Checks if output meets your standard. Pass or fix, with exactly what to fix.

Voice reviews · content audits · consistency checks

context organizer

Gathers scattered information into one brief

Pulls together everything needed before a task runs.

Session prep · weekly briefings · discovery call prep

gap researcher

Finds what's missing and fills it

Searches, analyzes, delivers a summary.

Prospect research · market analysis · content angles

03 · what you get

the thinking and the system.

AI configuration

Your methodology, voice rules, quality standards, and reasoning patterns. One file that teaches the AI how you think.

quality rubric

Scores every output before it reaches you. Does this sound like you? Pass or fix.

task instructions

Focused instructions for specific tasks. One for LinkedIn, another for client prep, another for content.

AI playbooks

The instructions that make AI run your operations your way. A Claude skill, a scheduled agent, or a full pipeline. New call booked? The system researches the prospect, drafts a prep brief, and emails it before the meeting.

live running system

Deployed and working. Not prompts in a document. Responds to real triggers in your business.

enterprise grade privacy

API level access. Your data, frameworks, and client information are never used to train public AI models.

inside your system

/your-core/
├── CLAUDE.mdyour AI configuration file
├── voice-rubric.mdquality scoring tool
├── skills/
├── voice-qa.mdchecks every output
├── linkedin-post.mdcontent in your voice
├── discovery-prep.mdpre-call research
├── playbooks/
├── session-to-content.md1 session → 5 assets
├── client-onboarding.mdnew client flow
├── agent.pythe orchestrator

what runs under the hood

Claude (the AI) · Claude Code (the build environment) · A server running 24/7 · An interface your team uses to talk to it.

ClaudeNext.jsTypeScriptPythonVerceln8nReactNotionClaudeNext.jsTypeScriptPythonVerceln8nReactNotion
04 · why it compounds

each process builds faster than the last.

By the second process, the AI already knows your voice and standards.

layer 1
first process
layer 2
second process
layer 3
third process
layer 4+
additional processes

let's figure out what to build.

It starts with mapping how your business actually runs. Then we decide what to build first. If it makes sense, we build. If not, you keep the map.

apply for CORE

let's see if it's a fit.

Two quick steps. I read every one.