02 · Working Intelligence Audit
I look at everything you’ve already taught your AI.
Fixed-scope · ~5–10 hours of work over one calendar week. Standard $497. With Bundle (export pack + 30-day Q&A) $997.
Anyone who already uses AI seriously and hasn’t yet thought about the asset they’re quietly building inside someone else’s house.
The patterns you’ve trained into Claude. The documents ChatGPT remembers. The projects Gemini knows. I extract it, structure it, and hand it back as a file you can actually own.
You’ve already built something. You just can’t see it yet.
Stop and look at your ChatGPT account for a second. The custom GPTs you’ve made. The conversations you’ve had a hundred times that taught the model how you write, what you care about, what you don’t want to hear. The Claude projects with the system prompt you’ve been refining for six months. The Gemini history that knows your last three quarters of work because it’s linked to your Google Workspace.
You’ve been training those models. You probably didn’t mean to. There was no moment where you sat down and decided to teach a frontier AI system how to think about your work. It happened the way a path gets worn into a lawn — one step at a time, until the line is unmistakable.
The thing you’ve trained is real. It’s an asset. If somebody copied it to another account right now, they’d have a significant head start on you in your own role. And right now it lives entirely inside OpenAI, Anthropic, or Google’s servers — in a format you can’t export, structured for their convenience, deletable any time they change a policy.
The Working Intelligence Audit is the systematic version of going through that asset, naming what’s there, and turning it into a file you own.
The diagnosis
Two ownership problems, compounding.
The first.Your AI training is locked inside three different vendor accounts. ChatGPT’s memory feature stores what it’s learned about you, but you can’t read it as a structured document. Claude’s projects let you upload files but don’t expose a clean export of the conversation patterns. The three accounts don’t talk to each other.
The second.If any of those vendors changes a policy, raises a price, deprecates a feature, or terminates an account — the asset goes with them. The fix is to extract it before you’re forced to.
“The asset you’re building inside Claude or ChatGPT is real, valuable, and quietly stuck in someone else’s vault. The Audit gets the key cut.”
What the Audit actually does.
I get supervised access to your AI accounts — read-only, on your machine, with you in the room or async with screen recording, your call. I work through them systematically using a structured extraction process I’ve refined across thirty-plus accounts.
Four layers come out of the extraction. Each one names something that was implicit before — patterns the model knows but you couldn’t describe out loud.
Four-layer extraction
I write the result up as a portable Markdown bundle — structured, named, organized in a folder tree you can drop into any modern AI tool and have the new model pick up where the old one left off. You walk away owning it. Day one. No subscription required to keep it.
The hygiene move
Back up the brain you didn’t realize you were building.
Five to ten hours of structured extraction, one written report, a portable corpus that's yours forever. Fee credits against the Install if you upgrade within 90 days.
Four things you walk away with
In the bundle.
The corpus
A folder of Markdown files organized for retrieval by any AI tool. Drop it into a Claude project, Cursor, ChatGPT — or the starting kit for a Personal Brain.
The Audit Report
3–5 page written read on what I found. What you've been building. What AI setup your role needs. Honest read — I'll tell you not to buy if the answer is no.
The 30-day Q&A
A month of follow-up email questions, answered substantively. The audit's value lands over weeks, not days — this is the window for it to compound.
The Install credit
Upgrade to a Personal Brain Install within 90 days and the Audit fee credits against it. You don't pay twice for the same thinking.
How a week of audit work runs
Five working days, plus the read-back.
- MON
Kick-off call
60 minutes. I show you the extraction process. We agree on scope — accounts, projects, timeframes. You set up read-only access.
- TUE–THU
Extraction
I work through your accounts methodically. Daily notes on what I'm finding so you can course-correct.
- FRI
Synthesis
I structure the corpus, write the Audit Report, assemble the bundle. You get a draft for review.
- MON
Delivery
Walk through the report together. Questions answered. You leave with the bundle + a clear read on next moves.
- +30
Q&A window
A month to apply what we found before any further decision. Email questions answered with substance, not drive-bys.
Standard ($497) vs. Bundle ($997).
Standard, $497. The corpus + Audit Report + 30-day Q&A window. The diagnostic at full depth. Most operators do this version and stop here — the corpus alone changes how you think about your AI setup for the next year.
Bundle, $997. Standard, plus: an export pack with three Claude project starter configs based on the patterns I found in your work, an installed-skill kit (3 role-specific reusable skills you can drop into your existing accounts), and a Personal Brain Quickstart — a basic version of the brain architecture, hosted on Cloudflare for the first 60 days while you decide whether to upgrade. The Bundle exists for operators who already know they want to go further; the credit-against-Install math means you almost always come out ahead.
If you’re not sure which to pick, start with Standard. You can always upgrade later if the audit surfaces something worth building on.
Who this is for
Named by role.
- The heavy-AI knowledge workerYou’ve been using ChatGPT or Claude as a daily collaborator for a year or more. You have memory turned on, custom instructions tuned, projects you’ve fed hundreds of files into. You want to see what you’ve actually built.
- The agency or studio ownerYour team uses AI across client work. There’s no central record of the patterns, prompts, or projects that have been refined across engagements. You want a structured read on the IP that’s drifting around in individual accounts.
- The career-transition operatorYou’re changing roles, leaving a company, or stepping into a new domain. You want to take your AI training with you — and you want to show a portable artifact to the people deciding whether to bring you on.
- The buy-vs-build decision-makerYou’re weighing whether to invest in a Personal Brain Install or stick with what you have. The Audit is the honest read — including the version where the honest answer is "you don’t need anything else."
What’s included
In the price.
- Read-only structured extraction across ChatGPT, Claude, and Gemini
- Four-layer map: domain encoding · workflow calibration · behavioral relationship · demonstrated capability
- Portable Markdown corpus, organized into a folder tree you own outright
- 3–5 page written Audit Report with a named recommendation
- 30-day post-audit Q&A window over email
- Kick-off call (Monday) + delivery call (the following Monday)
- Daily notes during extraction so you can course-correct
- Audit fee credited against Personal Brain Install if upgraded within 90 days
Common questions
Asked, answered.
Why pay for an audit at all — can’t I just export my own data?
You can export individual artifacts — Claude project files, ChatGPT conversation histories (in some cases), Gemini search results. What you can’t do is structurethe result. The raw export is useless without the analytical layer that turns "these are 800 chat messages" into "here are the four patterns those messages encode, and here’s how to use them in any future tool."
The Audit is the analytical layer. The extraction is the small part. The structure and the report are the work.
What if I’m worried about giving you access to my accounts?
Three options. Observed sessions — you screen-share, I direct, you take the action. Slower but you never hand over credentials. Screen recording — you record the relevant sessions on your end and send me the videos. Read-only temporary access— you create a temporary account-share via the vendor’s mechanism (where available) and revoke it the moment the audit ends.
I’ve done all three. The choice is yours; the result is the same. I treat the access window the same way I treat any sensitive engagement — work happens on your machine where possible, nothing is stored on my side beyond what makes it into your corpus.
What if the audit finds I don’t need a Personal Brain Install?
Then I’ll write that in the report. Most people who do this audit end up doing the Install eventually, but some don’t — usually because their work is already well-structured by the tools they have, or because their volume doesn’t justify the productized infrastructure. I’d rather give you the honest read than upsell you into something you don’t need. Your file is yours regardless.
Can I do the audit on just one account, not all three?
Yes. Single-account audits scope to the lower end of the time band (closer to 5 hours than 10) and the price comes down by the corresponding margin — talk to me about it in the kick-off call. The reason the standard offer covers all three is that most heavy AI users have meaningful patterns in each, and the highest-value insights usually come from comparing them — what Claude knows that ChatGPT doesn’t, what your Gemini history reveals about workflow that the others can’t.
What about Cursor, GitHub Copilot, or other code-side tools?
I can include them in scope if your work is engineering-heavy. The four-layer map adapts well to code contexts (the demonstrated-capability layer especially — what you’ve actually shipped with AI assistance is portfolio gold). Flag this in the kick-off call and we’ll talk about the right shape.
How long does the corpus stay useful?
Indefinitely. Markdown is the longest-lived format in computing — every AI tool that exists or will exist can read it. The structure I use is designed to outlast specific vendors. You’ll add to it over time as your work evolves, but the bones don’t age.
Is this confidential?
Yes. I sign whatever NDA you put in front of me. The work product is yours. I don’t train models on your data. I don’t case-study you without explicit written permission. The default is full confidentiality.
When would I schedule this?
The Monday kickoff is the anchor — I run one audit at a time so each gets full attention. Booking window is usually 2–3 weeks out. If you have a deadline (a job change, a board meeting, a quarter-close), say so and I’ll work to fit it. The booking link shows real availability.
Adjacent paths
Where this connects.
Book the Audit. Or the Diagnostic first.
Not sure if the Audit is right? Start with the 30-minute Diagnostic Call — it’s free, and it’s scoped to figuring out which offer fits. If the answer is "this one," we book the audit at the end of the call.
As featured in
