Build Your Brand Brain AI Setup: A Step by Step Guide

Most people use AI like a stranger at a counter. Every chat starts with "I run a small business that sells..." and ends with output that sounds like everyone else. A brand brain AI setup fixes that by giving the assistant your whole business once, in one place, so later work starts from what it already knows.

This guide walks through the setup from zero, in order. It follows Lesson 1.2, Build your brand's brain, which shows the full process on screen. Read this first if you want the map, then do the lesson with your own files open.

A brand brain is one folder plus a short set of notes. You gather what already exists about your brand, state one goal, organise it, sort your media, and end every chat by saving what was learned.

What a brand brain actually is

It is not a special product. It is a folder on your computer that holds everything an assistant needs to know about your business: who you sell to, how you sound, what you offer, what has worked, and what the current goal is. The assistant reads it, so you stop retyping it.

Tools like Claude Code are built around this idea. Anthropic's documentation describes a file called CLAUDE.md that holds instructions and context Claude should know every session, and says Claude reads it at the start of every session. It also describes auto memory, where Claude saves notes such as your preferences as it works. The brain you build here uses the same principle, applied to a business instead of a codebase. If you have not set up the tool yet, Lesson 1.1 covers that.

Step by step setup

1. Scan for what existsAsk the assistant to look through your computer for anything about your brand: old posts, price lists, logos, brand guides, customer emails, photos. Review what it finds before you keep anything.
2. State the goalWrite one line: views first, then followers, then sales. The order matters, because it tells the assistant what to optimise for right now.
3. Make one folderCreate a single folder for the whole business. Move or copy the useful finds into it, with plain names.
4. Sort the mediaSplit photos and video into three groups: ready to use, needs an edit, still to shoot.
5. Write the brain fileAsk the assistant to draft a short summary file from the folder: audience, voice, offers, goal. Edit it by hand until it sounds right.
6. Add the memory lineEnd every chat with one instruction that makes it save what it learned to the file.

Step one in detail: the scan

Be specific about where it may look. A good prompt names the folders and says what to look for. Something like: "Look in Documents and Desktop for anything about my brand. List what you find with a one line description, and do not change or delete anything." That last sentence matters. A scan is a read, not a clean up.

Expect noise. You will get duplicate logos, drafts you never published and files that have nothing to do with the brand. Discard freely. The aim is a short, true set of source material, not a complete archive.

Step two: the goal, in order

A brand without a stated goal gets generic advice. "Grow the business" gives the assistant nothing to choose between. The order of views, then followers, then sales is a sensible default for a small brand that is not yet visible, because people cannot buy from a brand they have not seen. Change it if your situation differs, for example if you already have an audience and need sales.

Put the goal at the top of the brain file, with a date. Goals change, and a dated line tells you when the file went stale.

Step three: one folder for the business

Resist elaborate structure. A simple layout is enough:

• brand: voice notes, audience, offers, the brain file itself
• media: the three sorted groups
• content: drafts and published posts
• results: notes on what performed

When the assistant works inside that folder, everything it needs is within reach. Anthropic's docs say that when you run Claude Code in a directory it can access the files in that directory and its subdirectories, which is why the single folder works.

Step four: sort your media

Open the media folder and ask for a three way sort: ready, needs an edit, still to shoot. The third group is the useful one. It turns "we need more content" into a shot list you can act on. The assistant can only judge files it can open, so check its sorting on a handful of items before you trust it.

Step five: the memory line

The brain only stays useful if it gets updated. Make it a habit to finish each chat with a line such as: "Save anything new you learned about my brand, audience or what worked to the brain file, and tell me what you changed." Read the changes. If something is wrong, correct it now, because wrong notes get repeated.

Check that it works

Test it with a fresh chat. Ask for a post about something small, with no background given. If the result uses your voice, your audience and your current goal, the brain is doing its job. If it reads generic, the file is too thin or too vague, so add one concrete example of writing you like and one list of phrases to avoid.

Where to go next

If you want a ready-made structure to fill in, use the free template. To avoid the usual wasted hours, read the five common mistakes. Then follow Lesson 1.2, Build your brand's brain to watch the whole setup done step by step, and come back to Lesson 1.1 if your tool is not ready yet.

Questions people ask

Do I need to code?

No. You describe what you want in plain words and review the result. The tool does the file work.

How long does it take?

The lesson is about 21 minutes of video. Your own setup depends on how much material you have to sort.

Does it replace my judgement?

No. It saves you retyping context. You still approve what gets published.