Every time you open ChatGPT, you start from zero, which is why learning how to brief an AI partner on your business matters. You explain who your customers are, what you sell, how you price, and why last month was a bit strange. Then you get an answer that could have been written for any business on the planet.
That’s not an AI problem. It’s a briefing problem. Nobody ever showed you how to give it a useful briefing, so it meets you for the first time every time.
This post fixes that. I’ll walk you through the brain dump session I run with founders. It takes one afternoon, a phone recorder, a handful of documents, and a running order that gets the knowledge out of your head and into files an AI reads every time you talk to it. You’ll know what to cover, what to skip, how to structure it, and how to test whether it worked.
Key Takeaways
- Knowing how to brief an AI partner on your business matters more than which tool you pick, because generic input always produces generic answers.
- A focused brain dump of about three and a half hours captures most of what an AI needs: offer, customers, decisions, team, numbers and strategy.
- Talk, don’t type. Speaking runs at roughly four times typing speed, so voice notes pull out detail founders skip when they write things down.
- The most valuable context is how you make decisions, not what you sell. Capture the judgement calls your team currently brings to you.
- Split the brief into small, plain-language files by topic so the AI loads what it needs, and you can update one area without rewriting the lot.
- Test the brief with real questions your team asked you last week. Every vague or wrong answer points to a gap in the context, not a flaw in the AI.
- A brief that never gets updated goes stale within a quarter. Ten minutes a week keeps it accurate as prices, people and priorities change.
Why Your AI Keeps Giving You Generic Answers
Picture hiring a sharp general manager. Smart, experienced, keen. On day one you give them a laptop and no induction at all, then you ask them to quote a job, handle an unhappy client and plan next quarter.
They’d do a decent job of sounding sensible. They’d also get half of it wrong, because they don’t know that you never discount for that client, that the cheap supplier burned you twice, or that the business is quietly moving away from small jobs this year.
That’s exactly where most founders sit with AI. The models are very capable now. What they lack is your context. Anthropic, the company behind Claude, has written about this shift and calls it context engineering: what you load into the model matters as much as how you phrase the question.
What lives only in your head
Ask a founder to list what their business knows and they’ll point to the website and maybe a folder of SOPs. The real knowledge sits elsewhere. It’s in the pricing exceptions, the clients who get priority and why, the warning signs that a job is about to go sideways, the way you word an awkward email.
None of that is written down. It comes out of you one question at a time, usually when a staff member catches you between meetings. That’s why your team can’t decide anything without you, and why the AI can’t either.
A brief is not documentation
You’ve probably tried writing it all down before. The SOPs got written, filed and never opened again. That’s the normal fate of documentation, because a document just sits there waiting for someone to read it.
A brief for an AI works differently. The AI reads it every single time, before it answers anything. It doesn’t skim, forget or get bored. Once the knowledge is in, it’s used on every question, which is why the brief becomes the brain of your business and not another dusty folder. If you want more on using AI this way, I’ve written about AI as a thinking partner and why the context you give it decides the quality of what comes back.
Before You Start: How to Brief an AI Partner on Your Business Without Wasting the Afternoon
The brain dump works because it’s concentrated. Fifteen-minute bursts between jobs produce a thin, scattered brief. One protected afternoon produces something you can use the next morning. A little preparation makes the difference.
Block the time properly
Put three and a half hours in the calendar and treat it like a client meeting you can’t move. Phone on do not disturb, email closed, someone else covering the front line. If you can, do it away from your desk. A kitchen table or a quiet meeting room keeps you out of the habit of checking things.
Gather your raw material
Spend thirty minutes the day before pulling together:
- Your website copy and any brochures or capability statements
- Your last three to five quotes or proposals, including one that won and one that lost
- Your current price list or pricing logic
- A list of your team with their roles
- The last three months of headline numbers: revenue, enquiries, jobs won
- Two or three emails you’re proud of, and one complaint you handled well
- Any SOPs or checklists that already exist, however rough
You don’t need to tidy any of it. The AI will sort it later. The point is to have real examples on hand, because real examples teach the AI far more than descriptions do.
Talk, don’t type
Record yourself. Use the voice memo app on your phone, a transcription tool, or the voice mode built into ChatGPT or Claude. Most people speak at around 150 words a minute and type at about 40. More importantly, when you talk, you explain your reasoning. When you type, you summarise and leave the useful bits out.
Get interviewed if you can
The best brain dumps happen when someone else asks the questions. A trusted team member works well, because they’ll ask about the things they actually get stuck on. If nobody’s free, ask the AI to interview you. Tell it you’re briefing it on your business and want it to ask one question at a time, follow up on vague answers, and push for examples.
One mindset rule for the whole session: talk as if you’re handing the business to a capable replacement for two weeks while you’re completely unreachable. What would they need to know so nobody calls you?
How to Brief an AI Partner on Your Business, Hour by Hour
Here’s the running order. It’s built so the easy material comes first, warming you up for the part that matters most, which is how you make decisions.
Block 1, 45 minutes: The business in plain English
Start with the basics, but go further than your website does. Cover:
- What you sell. Every product or service, including the ones you’d rather not sell any more.
- Who buys it. Describe your best three clients in detail. Then describe your worst three and how you’d spot them earlier next time.
- Why people choose you. Not the marketing line. The real reason, the one clients say on the phone.
- What you refuse to do. Jobs you turn down, clients you won’t take, promises you never make.
- How you think about price. When you charge more, when you’ll move, and when you walk away.
Founders rush this block because it feels obvious. It isn’t obvious to the AI. If it doesn’t know that you avoid small jobs under a certain size, it’ll happily draft a cheerful quote for one.
Block 2, 60 minutes: How decisions actually get made
This is the heart of the session and the part most briefs miss. Your offer is on your website. Your judgement isn’t anywhere.
Start by listing every question your team asked you last week. Most founders come up with fifteen to thirty once they think about it. Then work through them one by one and say out loud what you answered and, more importantly, why.
Some prompts that tend to pull out the good stuff:
- A client asks for a discount. What do you do, and does it depend on who they are?
- A job is running over time or budget. Who do you tell, and when?
- Two urgent things land at once. How do you decide which comes first?
- A supplier lets you down. When do you give them another go, and when are they done?
- A staff member makes a mistake with a client. How do you handle it?
You’ll notice patterns as you talk. “I always ring rather than email when there’s bad news.” “Repeat clients jump the queue.” “Anything over a certain value needs a site visit first.” These rules of thumb are gold. Say them plainly, and give an example of each one in action.
A tradie I spoke with recently realised the most important thing in his head was which quotes were worth chasing. He knew instantly from the address, the way the enquiry was worded and whether the client had asked about price first. Nobody else in the business knew any of it. Ten minutes of talking turned that instinct into a set of rules the AI could apply to every new quote.
Block 3, 45 minutes: People, numbers and rhythm
Now cover the machinery of the business.
Your team. Who does what, who’s strong at what, who to ask about which topic, and how each person likes to receive information. Keep it professional. This brief may end up shared, so leave out anything you wouldn’t say to their face.
Your numbers. Which figures you look at, how often, what good looks like and what makes you worried. “If enquiries drop below twenty in a week, something’s broken” is exactly the kind of line the AI needs.
Your rhythm. Weekly meetings, reporting days, busy and quiet seasons, the week before Christmas when everything goes sideways. Timing context stops the AI from suggesting a big marketing push in your slowest month.
Block 4, 30 minutes: Strategy and voice
Finish with where you’re heading and how you sound.
Strategy. What you’re trying to achieve this year, your top three priorities this quarter, and what you’ve decided not to do. The “not doing” list is underrated. It stops the AI from cheerfully recommending the exact thing you ruled out last month.
Voice. How you write to clients. Words you use, words you hate, how you open and close an email, how formal you are with different people. Then hand over those emails you’re proud of. Three real examples teach tone better than a page of description.
The last 30 minutes: Hand it over
Stop talking. Get the recording transcribed, which most tools do automatically, and gather the transcript plus every document you pulled together. That’s your raw brain dump. On most afternoons, it runs to somewhere between 15,000 and 25,000 words. Don’t edit it yet. The next step does the heavy lifting.
Turning the Brain Dump Into Context Files the AI Can Actually Use
A 20,000-word transcript pasted into a chat window is better than nothing, but not by much. It’s rambling, repetitive and hard for anyone to update. The final step turns the raw dump into a clean set of files.
Let the AI do the sorting
Give the AI the transcript and documents, and ask it to organise everything into separate files by topic. A good starting set looks like this:
- Business overview: what you do, who for, why you win, what you refuse to do
- Customers: ideal clients, bad-fit clients, warning signs, common questions
- Offer and pricing: every service, how you price, when you move and when you don’t
- Decision rules: the judgement calls from Block 2, each with an example
- Team: roles, strengths, who handles what
- Numbers: what you track, what good looks like, what triggers concern
- Strategy: this year’s goals, this quarter’s priorities, the not-doing list
- Voice: tone, phrases, sign-offs, example emails
Ask it to write in plain language, keep each file to one or two pages, and use short headings so it’s easy to scan. Ask it to flag anything it wasn’t sure about rather than guessing.
Review every file yourself
This bit is non-negotiable. AI is very good at filling gaps confidently, and some of what it writes will be slightly off. Read each file with a red pen mindset. Fix anything wrong, cut anything that’s no longer true, and answer the questions it flagged.
Put a date at the top of each file, such as “Current as of October 2026”. Pricing, team and strategy change. Dates tell you, and the AI, how fresh the information is.
Put the files where the AI reads them every time
Where the files live depends on how you use AI. If you’re on ChatGPT or Claude, both offer projects where you upload files that load into every conversation in that project. Create a project for your business, upload the files, and add a short instruction telling the AI to use them before answering anything.
For founders who want to go further, the same files can sit in a folder that an AI agent reads at the start of every session, alongside your live data and meeting notes. That’s when the brief stops being a reference and becomes the brain of a system that does real work. I cover where that leads in my post on AI agents for small business, and the agentic workflow post shows what happens once the brain starts handing tasks to an AI workforce.
Keep it small and specific
The temptation is to stuff everything in. Resist it. A tight two-page decision rules file beats a forty-page manual, because the AI weighs what matters more clearly when there’s less noise around it. If a file grows past three pages, split it.
Testing the Brief and Keeping It Alive
You don’t know whether the brief works until you test it with real questions. This is the part founders skip, and it’s the part that turns a nice exercise into something that changes your week.
Run the team-question test
Go back to the list from Block 2, the questions your team asked you last week. Put each one to the AI exactly as it was asked. “A client wants to push the job back a fortnight and keep the original price, what do we do?”
Score each answer: right, close, or wrong. Most first briefs land around half right, a third close and the rest wrong. That’s normal and fine. Every wrong or close answer tells you precisely what’s missing. Add that context to the right file and ask again.
Run the strategy test
Ask a bigger question: “Based on everything you know, what should my top three priorities be this quarter, and what should I stop doing?” If the answer names your actual team, your real numbers and the direction you described, the brief is working. If it sounds like a generic business book, the strategy and numbers files need more detail.
Run the hands-off test
Give one team member access to the project for a week. Ask them to try the AI before they come to you. Then count how many questions still reach you. The ones that do are the next gaps to fill. This is the first real sign the business is starting to run on a system and not on your memory.
Keep it current
A brief goes stale faster than you’d expect. Prices change, people leave, priorities shift, and within a quarter an untouched brief starts giving confident, outdated answers. That’s worse than no answer.
Two habits keep it alive:
- Ten minutes on Friday. Update anything that changed that week. New client rule, new hire, a lesson from a job that went wrong.
- The answer-once rule. Any time you answer a team question the AI couldn’t, add the answer to the brief straight away. You answer it once, and the AI answers it forever after.
Every quarter, reread the whole set and cut what’s no longer true. If you’re tracking the bigger picture of how AI is changing your workload, my 90-day AI automation checkpoint is a good structure for that review.
Conclusion
Briefing an AI isn’t a technical job. It’s the same induction you’d give a senior hire, done once properly and then kept current. One afternoon of talking, an hour of review and a few minutes a week is enough to turn a generic chatbot into something that knows your clients, your rules and your voice.
The bigger win is what it does to your week. Every judgement call you capture is one more question your team doesn’t need to bring you. That’s the first real crack in the trap most founders live in, where the business only works because you’re working it.
The brief is the brain. Once the brain is in place, the next step is an AI workforce that acts on it: answering enquiries, following up leads and handling the admin that buries most small business owners. That’s when the hours start coming back.
Ready to Get It Out of Your Head?
If you’ve read this and thought, “I know I should, I just won’t get to it”, that’s normal. Most founders are too busy inside the business to step out of it for an afternoon.
I’m happy to help you work out where to start. Book a free 30-minute Discovery Call, and we’ll look at what’s trapped in your head, what your team keeps asking you, and what a working brain for your business would look like. No pitch deck, just a straight conversation.
Frequently Asked Questions
What information should I give an AI about my business?
Start with what you sell, who buys it, and why they choose you. Then add how you make decisions, your pricing logic, your team and who does what, the numbers you watch, your current priorities and how you like to write. The decision rules matter most, because they are the part that lives only in your head and the part your team keeps asking you about.
How long does it take to brief an AI on my business?
A solid first version takes one focused afternoon, around three and a half hours of talking plus an hour to review the files the AI produces. That covers most of what it needs to give useful, specific answers. After that, ten minutes a week keeps it current. Trying to do it in scattered fifteen-minute bursts usually takes longer and produces a thinner brief.
Is it safe to share business information with an AI tool?
It depends on the plan and settings you use. Business and team tiers of the major tools generally do not train on your data by default, while some free and personal plans may, unless you switch it off. Check the data controls before you start, leave out passwords and banking details, and keep client personal information to what the AI genuinely needs.
Which AI tool is best for holding my business context?
Any of the major assistants can hold a brief. ChatGPT and Claude both offer projects where you upload files that load into every conversation in that project. The tool matters less than the quality of the context you feed it. Pick the one you already use, put the brief in properly, and only switch if it genuinely struggles with your files.
How often should I update my AI’s business context?
Do a quick update every week and a proper review every quarter. Weekly, add anything that changed: new prices, a new hire, a client rule, a lesson from a job that went sideways. Quarterly, reread every file and cut what is no longer true. A brief that sits untouched for three months starts giving confidently outdated answers, which is worse than no answer.
Can my team use the same AI brief I created?
Yes, and that is where most of the value shows up. When your team can ask the AI how you would handle a pricing exception or a difficult client, they stop routing every question through you. Share a team version that leaves out anything private, such as your personal goals or sensitive staff notes, and keep a fuller version for yourself.
About Octavius
Titus Mulquiney is the founder of Octavius AI, where he builds AI brains and AI workforces for founder-led businesses stuck running everything out of their own head. Twenty years in marketing, ex-Sony product manager, ex-GM Zeal NZ. Based in Auckland, working with operators across NZ, Australia, and the US. Connect on LinkedIn.