How to Build an AI That Actually Knows Your Business
Every time you open ChatGPT, you start over. The AI has no idea who you are, what you do, how you price jobs, or what your clients expect. You type a question, get a generic answer, then spend 10 minutes typing in context it should already have. Most business owners accept this as just how AI works.
It doesn't have to be that way.
There are features built right into ChatGPT, Claude, and Google Gemini that let you give AI permanent context about your business. Most people haven't set them up. The ones who have are getting answers that are two to three times more useful — without writing longer prompts or doing anything differently — because the AI already knows the background.
Here's exactly how to do it.
The Context Tax
Every generic prompt you write costs you something. You ask the AI to draft a follow-up email and get something that sounds like it came from a Fortune 500 corporation. You ask it to write a proposal and get a template so generic you spend 30 minutes making it sound like you. So you add context: "I run an 8-person landscaping company in Utah, our tone is direct but friendly, our average job is $4,500..."
That's the context tax. And most business owners are paying it every single day.
I've worked with enough business owners to know that the people who feel like AI isn't really working for them are usually doing exactly this — pasting in fresh context every time, getting mediocre output, feeling vaguely disappointed. They're not doing anything wrong. They just haven't been told there's a better setup.
The answer isn't a smarter prompt. It's a smarter foundation.
What AI Memory Actually Is
A few years ago, every AI conversation was stateless. Close the tab, everything's gone. That's changed significantly.
ChatGPT now has persistent memory across conversations. It can remember what you tell it about your business and apply that context automatically to future conversations. Claude has a feature called Projects — you attach context documents to a dedicated workspace, and every conversation in that workspace reads from those documents automatically. Google Gemini has Gems. Microsoft Copilot has similar features inside Teams and Outlook.
These aren't perfect. They won't replace a real knowledge base or a well-built system. But for the average service business owner writing emails, proposals, client updates, and internal communications — they're enough to eliminate 80% of the re-entry work. And almost nobody is using them.
Build a Business Brain Document
Here's the approach I walk through with new clients, regardless of which AI tool they use: build what I call a Business Brain document. This is a single reference file — two pages, Google Doc or Notion — that captures everything an AI would need to know to actually help you.
Here's what goes in it:
- Who you are: Business name, what you do, how long you've been at it, where you're located, how many people are on your team.
- Who you serve: Your typical client — industry, size, the problems they usually come to you with, what makes them a good fit.
- How you price: Not exact rates, but your philosophy. Are you premium? Mid-market? How do you structure quotes?
- Your voice: How you write. Formal or casual? Short and direct or detailed? Phrases you use. Phrases you'd never use.
- What you don't do: Things that would misrepresent your business. Commitments you don't make. Services you don't offer.
- Current priorities: What's the business focused on right now? New service line? Specific market? Cutting a cost center?
This document doesn't need to be comprehensive. Two focused pages beats ten rambling ones. The goal is giving the AI enough to stop being generic — not writing an encyclopedia about your business.
The Setup: Tool by Tool
Once you have the document, here's how to plug it in.
ChatGPT: Go to Settings → Personalization → Memory and make sure it's turned on. Start a conversation and say: "I want you to remember the following about my business." Then paste in your document. ChatGPT will confirm it's saving each piece. You can view and edit what it knows anytime by going back to that Memory settings page. Over time it also picks up context from your conversations naturally.
Claude Projects: In the Claude web app, create a new Project. Inside, there's a section called Project Knowledge — paste your Business Brain document there. Every conversation you start inside that Project has access to it automatically. No re-pasting. The AI reads it with every message you send.
Google Gemini Gems: In Gemini Advanced, create a custom Gem with your business instructions and context. Save it. Access it any time you need AI help that should know your background.
Most business owners use one primary AI tool. Pick the matching setup, spend 30 minutes on it, and you're done. That's a one-time investment that makes every AI interaction afterward materially better.
What This Looks Like in Practice
Let me give you a real example.
I worked with a 6-person marketing agency. Before this setup, the owner was writing prompts like: "Write a proposal for a new social media client. We're a boutique agency focused on ROI over vanity metrics. Our packages start at $2,500/month. We work primarily with home services companies..."
Every single time. Same background, every prompt.
After setting up a Claude Project with her Business Brain document, her prompts became: "Write a proposal for a home remodeling company, in business about 10 years, wants to grow local presence."
That's it. The AI already knew the agency's voice, pricing philosophy, service structure, and how they position against larger agencies. Output went from a template she'd heavily edit to something she'd send with minor tweaks. She estimated it cut proposal time by about 40 minutes per proposal. They write 8–10 proposals a month.
That's 5–7 hours a month saved. From a 30-minute setup.
Who Gets the Most Out of This
This matters most if you write a lot of similar content for different clients — proposals, follow-ups, client updates, emails, internal briefs. The more your business runs on written communication, the more the context tax is costing you.
The businesses I've seen benefit most:
- Service businesses with repeat clients — You have ongoing relationships and a consistent voice that should carry across everything. Once the AI knows that context, every piece of communication gets better.
- Agencies and consultants — You're writing a lot, often in a specific voice for different clients. A Claude Project per client means the AI knows each one's context separately. No more accidentally using the wrong tone for the wrong account.
- Small teams where multiple people use AI — When everyone works from the same Project or shared workspace, your AI output is consistent across the team. You stop getting 10 different writing styles depending on who drafted the message.
- Business owners who delegate writing to AI — If you're using AI to produce content at volume, the quality ceiling is determined by how much context the AI has. More context, better output, less editing.
A Few Things This Isn't
This isn't a magic fix. The AI will still be wrong sometimes. You'll still need to review output before it goes out. A Business Brain document removes friction — it doesn't replace judgment.
On privacy: I get this question a lot. Here's the short answer — don't put in anything you wouldn't write on a sticky note on your desk. Business name, service description, pricing philosophy, general client types: fine. Social security numbers, confidential client financials, legal matters in progress: not in here. If your business handles regulated data, get proper advice before building this out in a consumer AI tool.
Also, this isn't a one-and-done thing. Your business changes. Your priorities shift. Revisit the document every quarter and update it. The more current it is, the better the AI output gets.
What's Coming in the Next Year
Right now, AI memory is something you configure intentionally. You decide what the AI knows. That's going to shift.
The major AI labs are building toward systems that learn from your interactions automatically — not just what you tell them, but patterns they observe over time. Which drafts you approve. What edits you consistently make. What types of questions you ask most often. The AI builds a working model of how you operate, not just the facts you've shared.
This is in early form already in some tools. Within 12 months, it'll be standard across the major platforms.
The business owners who've gotten their AI properly configured now — who have a Business Brain document, who've built out Projects or memory configurations — are going to compound those gains faster when this next wave hits. The AI will have richer context to learn from. Good setup pays forward.
The ones still typing generic prompts into a fresh chat window every morning are going to keep getting generic answers. And spending an extra 20 minutes on every task that should take five.
The 30-Minute Setup Worth Doing This Week
Here's what I'd actually do:
- Open a Google Doc. Write two pages about your business: what you do, who you serve, how you price, how you sound, what you never do.
- Pick your primary AI tool. Follow the setup steps above for that tool.
- Paste in your document. Do one test — ask the AI to write something you'd normally write yourself and see how close it gets without you adding extra context.
- Adjust the document based on what's missing. It takes a few iterations to get it dialed in.
No special tools. No new subscriptions beyond what you're probably already paying for. Just 30 minutes that makes every AI interaction afterward materially better.
This is one of the first things I set up with new clients. Not because it's flashy — it isn't. Because it's the foundation everything else builds on. You can't get good AI output without good AI context. Start here.
If you want help building this out for your specific business — or figuring out what the larger AI workflow should look like — book a free call. We'll figure out exactly what your setup should look like and what it's worth to your operation.