Build a second brain for your business
A notebook your AI keeps up to date, so it knows your business a little better every week. What it is, and how to start tonight with any AI app.
October 28, 2026 · By Bohol AI
In the October 28 Daily AI, the 401 tip was to build a second brain for your business. This guide explains what that means and how to start one tonight with the AI app you already use: ChatGPT, Claude, Gemini or NotebookLM.
A second brain is a notebook your AI keeps up to date for you. It holds what you know about your business: prices, hours, delivery rules, and the replies that work. Every week it gets a little better, so you stop starting from zero every time you ask.
Why it matters
- Faster replies you can trust. The AI answers from your own notes, so it stops guessing prices or stock, and customers get an answer before they message another shop.
- Your business isn’t only in your head. A helper or family member can answer customers when you’re busy, from the same notes.
- It keeps getting better. One short update a week keeps the notes current, and each update makes the next answer better.
Upload and ask vs. a second brain
Left: the AI digs through your files for every question, then forgets. Right: every answer builds on notes that keep getting better.
Most AI apps that take files work like the board on the left. You upload a pile of files, and every time you ask, the AI searches the pile again. Nothing is kept between questions, and if two files disagree, it may pick the wrong one.
A second brain adds one habit: once a week, the AI rewrites your notes with what it learned, and you check them. It then answers from notes that are already up to date, instead of a growing pile of files.
Your shop’s second brain
Step 3 is the one that turns a folder of files into a second brain.
- Collect: your price list, delivery rules, supplier contacts and the replies that worked best.
- Keep in one place: have the AI turn them into one “Business notes” document inside a project.
- Update every week: give it the week’s chats and changes. It rewrites the notes and flags anything that conflicts. You check, then swap in the new version.
- Ask, then check: ask it anything about your business. You still read every reply before you send it.
How to start tonight
One project, one notes document, one update a week.
- Make one project: a ChatGPT Project, Claude Project, Gemini Gem or NotebookLM notebook. Name it after your business.
- Write one “Business notes” document with your prices, opening hours, delivery rules, payment options and your five best replies. Only have a photo of your price list? Upload it and ask the AI to type it into the document.
- Paste Prompt A into the project’s instructions, so every chat follows it.
- Every week, use Prompt B with the week’s chats. Check what changed, then replace the old notes with the new version.
- Once a month, use Prompt C to find what is out of date or missing.
Prompt A: instructions
You help me answer customer messages for [business name]. Answer only from my Business notes. If the answer isn't there, say "Let me check and get back to you." Never make up prices, stock or delivery times. Reply in the customer's language (Bisaya, Tagalog or English), short and polite.
Prompt B: weekly update
Here are this week's customer chats and changes: [paste]. Update my Business notes: change any prices or rules that changed, add the common questions with answers, and list anything that contradicts the notes so I can check it. Give me the full updated notes.
Prompt C: monthly check-up
Read my Business notes. What is out of date, unclear or missing? What do customers ask that these notes can't answer yet?
Keep it safe
- Leave customers’ phone numbers and addresses, passwords and bank details out of your notes.
- One notes document that is right beats ten files that disagree. Add only what customers actually ask about.
- You still read every reply before you send it.
Where the idea comes from
AI researcher Andrej Karpathy calls this an “LLM wiki.” You collect the sources and ask the questions. The AI does the boring part: keeping the notes organized, linked and up to date, which is the work people usually give up on. Companies use the same idea for a whole team, with a person approving each fact before the AI can rely on it.
For the interactive whiteboard version of this guide, and a technical version for people building one with code or for a team, see the links below.
Sources
- Interactive whiteboard version of this guide
- Technical version: how LLM-maintained knowledge bases work
- Andrej Karpathy, “LLM Wiki” (the idea this guide is based on)
- Daily AI, October 28: Run your business messages