The short answer: You do it through structured conversation, not a questionnaire, and you do it in rounds. A founder cannot sit down and write out everything they know, because most of what they know does not feel like knowledge to them. It feels like common sense. The job is to ask the questions that make the invisible visible, record the answers, organize them into categories a stranger could use, and then come back for a second pass once the founder sees what they missed.

It is slower than a brain dump and it is the only method I have seen work.

Why the brain dump fails

Every founder has tried it. Block off a Saturday, open a blank document titled "How We Do Things," and start typing. By noon there are four pages. By the following week the document is forgotten, and six months later it is wrong.

The problem is not discipline. The problem is that expertise hides from the person who has it.

A founder who has handled a thousand customer calls does not experience that as a thousand data points. They experience it as instinct. Ask them how they qualify a lead and they say "I just know." Ask them why they never take projects under a certain size and they say "those never work out." Both answers are true and neither one is usable by anyone else.

When I was a Hospital Corpsman, the Navy taught with one rule: see one, do one, teach one. The teaching step is there because you do not actually understand a procedure until you have to explain it to someone who has never done it. The founder's knowledge works the same way. It only becomes transferable when someone makes them explain it.

The method: interviews in rounds

Here is what I do with clients. It usually runs three to five sessions over a few weeks.

Round one: the story. I ask the founder how the business started, what went wrong early, and what they would tell themselves on day one. I am not after a biography. I am listening for the moments where a rule got formed. "After that client, I stopped doing fixed-price work." That sentence is a policy, and it came out as a story.

Round two: the customer. Who is the ideal client? Describe the last one you loved working with. Describe the last one you fired. What did they say in the first call that told you which kind they were? What are they afraid of when they hire someone like you? What do they always ask?

Round three: the work. Walk me through a project from first contact to final invoice. Where do you personally have to be involved? Where does it stall? What do you check before anything goes out the door?

Round four: the voice. Show me three pieces of writing that sound like you and three that make you wince. What words do you never use? How do you open an email to a new prospect? What is the thing you say in every sales conversation that you have never written down?

Round five: the gaps. By now the founder has read the organized notes from the first four rounds and has a list of "oh, and also." This round is almost always the richest.

I record every session and have it transcribed. Then I sort what came out into categories.

The categories that matter

The organized output is the foundation layer of Brand Voice AI, and it looks like this:

  • Identity: mission, values, origin, what you refuse to do
  • Audience: avatars, fears, triggers, the questions they ask
  • Offers: what you sell, how you describe it, what it includes and excludes
  • Process: the steps, the checkpoints, the places a human must decide
  • Voice: tone, vocabulary, structure, do's and don'ts, example pieces
  • Stories: the anecdotes that teach a rule better than the rule does
  • Standards: what must be true before anything is published, sent, or delivered

A founder's knowledge almost always lands in all seven. If a category comes up empty, that is a finding too. A business with no written standards is a business where quality depends on who is having a good day.

The 674-file lesson

The reason I build it this way is that I did it to myself first.

In July 2026, I pulled 674 files out of a ChatGPT workspace I had been using for over three years. Marketing plans, drafts, research, half-built frameworks, conversation logs. I had been doing exactly what every founder does: having the thinking, then moving on without organizing it. The thinking was not lost, but it was useless, because it was scattered across hundreds of files with names like "Untitled 47."

Sorting those files into the seven categories above took weeks. Along the way I found rules I had forgotten I had, stories I had told dozens of times but never written down, and three versions of my own positioning that contradicted each other. By the end, I had something another person could read and understand, and something an AI tool could read and work from.

That sorting process became the interview method. Instead of waiting until a founder has 674 files of chaos, I go get the knowledge while it is still in their head and put it in order on the way out.

The veteran who could not sell

The story I tell most often is about a fellow veteran who spent years building a service business and decided it was time to sell. He had revenue, repeat customers, and a reputation. Buyers looked at it and all said a version of the same thing: "You are the business."

Every process ran through him. Every customer relationship was his. Every decision about pricing, quality, and who to take on lived in his judgment, and his judgment was not written anywhere. A buyer would be paying for a company that stopped working the day he left.

He did not have a bad business. He had an undocumented one. The knowledge was real and valuable and completely trapped. Capturing it earlier would not have changed what he knew; it would have changed what he could sell.

Most founders will never sell. The problem still applies. Every hire you train, every assistant you hand work to, every AI tool you try to use is a buyer of your knowledge, and they are all getting the same answer that buyer got.

What to do with it once it is out

Captured knowledge that sits in a document is only slightly better than knowledge in a head. The point is to use it:

  • New hires read it in week one instead of absorbing it over a year
  • Your assistant or contractor checks it before asking you
  • AI tools are given it as context, so their drafts sound like your business instead of a generic one
  • You read it yourself once a quarter and notice what has changed

That last one matters. Knowledge capture is not a one-time event. The second pass is always better than the first, and the business keeps learning.

A simple first exercise

If you want to start tonight without a consultant, try this. Open a voice memo app and answer one question out loud for ten minutes: "What do I know about my customers that nobody else in my business knows?" Do not edit. Do not organize. Just talk.

Then get it transcribed and read it. You will find at least three rules you have never written down. That is the method, scaled down to one round.

If you would rather have someone run the full process with you, that is what a Brand Voice AI foundation session is. And if you are not sure how much of your business is trapped in your head, the Business Systems Assessment is a seven-minute gut check on exactly that question.