Record yourself talking about how your business actually runs, turn those recordings into plain text files (your pricing logic, your processes, your way of saying things), and load the files into an AI workspace like a Claude or ChatGPT Project. From that point on, the AI answers like your business instead of like the internet. Done right, it costs about 30 minutes of talking a month. And once it's written down, it never quits, never forgets, and never walks out the door.
We audit AI visibility for a living, and this July we ran 40 live queries across ChatGPT, Perplexity, Google, and Google's AI Overviews. This exact question came back empty on all four engines. Nobody owns the answer. So here it is, the whole method, including the do-it-yourself version.
Why doesn't ChatGPT already know my business?
ChatGPT is brilliant about the world and clueless about your business. It has read most of the internet. It has never read your price list, your service-call policy, or the way you talk to a customer who's upset. So when you ask it to write a quote or answer a lead, you get something generic, and you end up rewriting it yourself at 10pm.
The problem isn't the AI. The problem is that everything that makes your business yours lives in your head and maybe one employee's, and nobody ever wrote it down where an AI could use it.
What business knowledge should I capture first?
Six things, in this order:
- Your pricing logic. Not the price list. The rules behind it: when you add 10%, when you give a break, when you walk away.
- The questions customers ask constantly. And the exact answers you give, in your words.
- How the work actually gets done. The handful of processes that matter, the way you'd explain them to a new hire on their first day.
- The jobs that went sideways. What happened and what it taught you. This is the most expensive knowledge you own, and it's usually stored nowhere.
- Your voice. The phrases you always use. The things you'd never say.
- Your boundaries. What you say no to. What you promise, and what you never promise.
If your best person gave notice tomorrow, this list is what would walk out with them. That's why it goes on paper first.
What's the fastest way to capture business knowledge for AI?
Talk. Don't type.
Owners have been buying process-documentation software for decades, and most of those binders sit unopened. Nobody builds their business a brain by writing documentation at night. Here's what actually works:
- Open a voice memo and answer the six items above out loud, like you're training a sharp new hire. Thirty minutes.
- Transcribe it. Your phone or any free tool does this now.
- Have AI turn the transcript into organized, plain-language files: one for pricing, one for process, one for voice, and so on. Ask it to flag anything unclear so you can fix it in two minutes.
This is exactly how we run it for clients. One of our realtor clients gives us one recorded call a month, about 40 minutes of her just talking. That single conversation becomes two publishable articles, an updated playbook, and a video script. She never types a word.
You talk. Something else writes. That's the only version of knowledge capture that survives contact with a busy week.
Where should the knowledge live?
In plain text files you own. Not locked inside somebody's software.
A folder of readable files does three things for you: you can open and check it, you can hand it to any AI tool (today's and whatever ships next year), and you can take it with you if you ever change platforms. If a vendor tells you your business's knowledge has to live inside their system, that's their moat, not yours.
Our own company runs this way. Clients, pipeline, pricing decisions, lessons learned: all plain files. The AI that runs our mornings reads them the same way you're reading this page.
From that folder, load the files into an AI workspace (a Claude Project or a ChatGPT Project both work). Now every conversation starts already knowing your business.
How do I put the knowledge to work once it's captured?
Two stages.
Stage one, day one: the chat that knows your business. Ask it anything: "what did we decide about weekend calls," "draft the reply to this customer," "price out this job using our rules." It answers like your sharpest ops person. New hires ask it before they ask you.
Stage two: AI doing jobs on its own. Leads answered in minutes, in your voice. Quote follow-ups that never die in silence. Admin drafted from a voice note. Our own system writes a brief at dawn every day: what's due, what slipped, what to say to whom. This is when it stops being a tool and starts being a department, and it only works because the knowledge underneath is yours. A generic bot answering your phone sounds like a bot. One running on your playbook sounds like your business.
Can I do this myself?
Yes. A focused weekend gets you version one, and the steps above are the whole recipe. Two honest warnings from doing this professionally:
- The value is in the rhythm, not the setup. A knowledge base built once is a snapshot that starts aging immediately. The businesses that get the compounding effect put 30 minutes of new knowledge in every month, and the system gets smarter every time.
- Stage two is real work. Wiring AI into your actual phone, inbox, and follow-ups, and keeping it current while the AI world changes every few months, is where most do-it-yourself projects stall.
Either way, start capturing. Every month the knowledge stays only in your head, you're paying for it.
Frequently asked questions
How long before it's useful?
The first hour. The moment your pricing logic and voice are loaded, drafts stop being generic.
Is my knowledge training someone else's AI?
Use a business-tier account and check its data controls, and keep the source files in your own folder. Plain files mean you can leave any platform with everything.
What if my business changes?
That's the point of the monthly rhythm. A 30-minute deposit keeps the system current, which is exactly what a binder printed in 2019 never did.
Do I need to be technical?
If you can talk and drag a file into a window, no. The technical work shows up at stage two, when AI starts doing jobs on your real phone and inbox.
If you'd rather have it done for you
Start with one win. If it doesn't work, you don't pay.
This is literally what we do: capture how your business runs, build the AI systems on top, and stay on as the team that runs them. The way in is a $1,000 Win-First Project: one agreed AI result plus a prioritized 90-day roadmap, and if the agreed result doesn't work, you don't pay.
Book a 20-minute fit call