Why AI gives you generic answers
You ask an AI something about your business and get an answer that sounds right. It is well-written. It just does not know anything about you.
It does not know your refund policy. It does not know how you handle a cancellation at 7 PM on a Saturday. It does not know what your team decided last Tuesday or how you actually run intake at your busiest location.
That is not the AI being bad at its job. That is the AI not having any of your material to work from.
The quality of what you get out depends almost entirely on what you put in. Same question, same tool, completely different answer once it knows how you actually operate.
Put the real stuff somewhere it can read
You do not need a perfectly organized knowledge base. You need the material you already have, in one place. Here is what that looks like:
SOPs and how-to docs
How you handle a cancellation. How you open. How you run intake.
Meeting notes
What was decided and why. Decisions are easy to lose and expensive to re-make.
Policies
Refunds, memberships, comps, cancellations. The things people ask about every week.
Onboarding material
What a new front desk hire needs to know in their first two weeks.
Anything you explain more than twice
If you have said it three times, it belongs in writing.
A simple test: if you answer a question by forwarding an old email or repeating yourself in Slack, that answer belongs in here.
This does not need to be tidy or complete to be worth doing. Half of it is better than none of it.
Why shared beats personal
Most people who use AI at work paste some context into a chat, get their answer, and move on. That context disappears when the chat closes. The next person starts from nothing.
Put your material in a shared space instead, and that context is there for everyone, all the time. Each person asks their own questions of it.
Picture the same Notion space, one afternoon:
A GM asks their AI to draft a reply to a member complaint. It already knows your refund policy.
An operations lead asks what was decided about the new intake process. The answer is in the meeting notes.
Someone building next month's schedule asks how the last one was structured. It is right there.
Nobody interrupted anyone. Nobody re-explained anything.
This is not really about AI. It is about your business knowledge existing somewhere other than inside people's heads. Connecting AI to it is what finally makes that worth the effort.
Same material, any AI
One person on your team uses Claude. Another uses ChatGPT. Both can read from the same Notion space. No one has to switch tools or share logins.
Connecting your AI
Most AI assistants can now be given access to read a Notion space directly. You stop copying and pasting material into every chat.
This works through something called MCP, which is an agreed-on way for AI tools to connect to other software. It matters for one reason: everyone does not have to use the same AI. One person can use Claude, another can use ChatGPT, and they are both reading the same Notion pages.
Claude
In Claude, go to Customize → Connectors and find Notion in the directory. Click Connect, then sign in to Notion and choose which pages Claude is allowed to see. That is the whole setup. You do not need to paste a URL or edit a config file.
Available on Claude Pro, Max, Team and Enterprise plans.
Full steps on Notion's developer docs →Before you connect anything
An AI can see whatever the person who connected it can see. Connect it to your whole workspace and it can read your whole workspace, including anything you forgot was in there.
Start with one space. Make a dedicated space for the shared material, put only that material in it, and invite people to that space.
Most business accounts let an administrator control which AI tools are allowed to connect at all. That is worth asking about before you roll this out to a team.
The access rule
- •Create one dedicated Notion space for shared material.
- •Put only that material in it. Nothing else.
- •Invite people to that space, not your whole workspace.
- •Connect AI to that space only.
- •Ask your admin which AI tools are approved to connect.
Start with five pages
If you are not sure where to begin, start here. Five pages covers most of what an AI gets asked about on any given day.
1. Your cancellation and refund policy 2. How you handle a new lead, start to finish 3. Notes from your last leadership meeting 4. Your opening and closing checklist 5. The thing you explained twice this week
Add pages as questions come up. The list builds itself if you let it.
Worth doing either way
This is something we have been doing and have found genuinely useful: the material is worth having even if nobody ever connects an AI to it.
Connecting the AI is what makes people actually keep it current.
If you want to go further with AI context, Stop re-explaining yourself to AI covers how to save your business context once so every answer comes back in your voice.