Live AI, screen shared.

Nothing starts from a slide.

It starts from something your team already made.

A sales call with the objection still in it.

A coaching session where the framework finally landed.

The Slack thread where somebody explained the problem better than any prompt could.

We put that on the screen and build a workflow out of it while everyone watches.

The person who's going to use the thing is the person who builds it.

Which is most of why it's still alive the week after I leave.

Some rooms don't get there.

If a team shows up with nothing of its own to work on, the session collapses into a demo.

And a demo is worth about as much as the notes people take during it.

So I ask for the ugly material up front.

Prompts age out in a month.

What holds is how you think in front of the machine.

And in every room it comes down to the same few habits.

1. 01

   Start at the end state

   Describe what finished looks like before you ask for anything at all.

   Everything after that gets easy.

   Because now you're only asking what has to be true to get there.

   “Now it becomes actually really simple. Because now all we have to do is say... what are the steps between the start and the end state?... What are all the things that must be true for this to happen?”

2. 02

   Catch the answer that sounds right

   This is the one that costs teams the most.

   Because it feels like progress right up until somebody tries to use it.

   It has a name in the room.

   “It sounds really good. But for AI... it is what I call directionally right. Operationally. Useless. Because it's not telling AI exactly what is the next thing it needs to do.”

   It's one of the most dangerous parts of AI.

   And the danger is that it feels like a win.

   The answer sounds so good that everyone relaxes.

   And the trouble only starts when somebody tries to put it in a deck or actually run it.

3. 03

   Make it show the plan before it does the work

   One line, typed before anything else.

   And it changes what comes back.

   “Before you do this, please explain to me the steps you will need to take to get there.”

4. 04

   Then make it show proof

   AI's incentive is to keep you on the platform and using it as much as possible.

   So AI is going to try to make you happy. As fast as possible.

   Which means it'll give you what sounds good instead of what's right.

   “Give your step-by-step plan. And when you execute this, show me proof that you did it.”

5. 05

   Fix the cause, not the output

   A bad result gets corrected and everyone relaxes.

   The workflow that produced it is untouched.

   So it produces another one next week.

   “It's going to force you to do the step before, or try and figure out what is the root cause of whatever's going on. Otherwise you're going to continue just to put Band-Aids over symptoms.”

Underneath all five is one idea people find mildly insulting the first time they hear it.

Most teams assume they have to go make something new before AI can be useful to them.

They're sitting on the material already.

And they've never once looked at it as material.

Pulling that out of a person and writing it down is the work I do under the name Cognitive Fingerprint™.

A workshop is the compressed version.

We take one call the team already recorded, get the thinking out of it, and turn it into something they can use on Monday.

Four of them, depending on what the room needs.

All four run on your material, so no two deliveries look the same.

1. 01

   Team AI operating

   Shared habits for a team that isn't going anywhere near code.

   We map where the work actually lives, build the workflows that keep repeating by hand, and leave a resource hub they'll come back to.

2. 02

   Claude Code setup

   For operators who want their own working setup, with memory, file structure, skills and verification wired in.

   Plus the discipline that stops the whole thing rotting in six weeks.

3. 03

   Transcript to asset

   For teams sitting on calls, trainings, podcasts and coaching sessions.

   We turn one into something real in the room, then reverse-engineer how we did it so they can do the next one without me.

4. 04

   Build day

   A site, a dashboard, an internal tool.

   We name the annoyance, cap the features hard, plan it, build the core, verify it, ship it.

The session doesn't end with "good luck."

You keep the recording and the transcript.

The session notes. The prompts.

The workflow instructions written down in plain English.

A resource hub or workspace.

A map of what happens first.

If we built something live, that's yours too.

What I can't hand you is the habit.

That gets built by the team using the thing for the three weeks after I'm gone.

And some teams don't.

The ones that do are the ones where somebody in the room did the building instead of watching me do it.

The Claude Code Accelerator.

The AI Brickyard.

Four sessions for one client's team, from what's-possible through to their own working second brain.

A weekly AI builders club inside a private founder community.

The format runs from a single talk to a hands-on day to a cohort that meets every week.

Best for teams that touch AI every day and keep getting the average answer out of it.

Agencies, coaching teams, service businesses, creator-led companies.

And founders who want AI inside the business rather than floating somewhere near it.

Bring the workflows you're embarrassed by.

Those make the best sessions.