A company brain is a single place that holds what your business knows: who your people and clients are, what was said in meetings, what was decided and why. It is kept in a form that both your team and your AI agents can use. We built one for Hawkfry and now run the business on it. This is how it works, what it runs on, and what we learned along the way.
Why AI Needs to Know Your Business
Most of the businesses we work with want AI to do more for them, and most of them hit the same wall: the AI doesn't know anything about the business.
Some of what it needs is easy to reach. The CRM has the deals and the accounts system has the invoices. The hard part is everything in between. Why did we take that approach with a client? Who do we actually know there? What did we agree on that call in September? Those answers are spread across meeting recordings, email threads and people's memories.
We were no different. We had hundreds of recorded client meetings going back years, and when one of us needed something from them, the quickest route was still to ask whoever had been in the room.
The obvious fix is to hand an AI all the transcripts and let it search. In practice that is slow and expensive, because the AI has to read far more than it needs. It also gets things wrong, because a transcript records what people said, which is not the same as what was decided.
How We Built Our Company Brain
We organised what Hawkfry knows into a set of linked notes that our team and our AI agents share. There is a note for every meeting, person, client, project and tool we use, and they link to each other the way pages on a website do.
That really is the whole system. We didn't buy a platform. We keep the notes in a consistent shape and follow a few rules about how they are maintained.
Keep the Record Separate From the Summary
When a meeting ends, the write-up of that meeting is saved and never edited again. That gives us a faithful account of what was said at the time.
Separately, we keep one living note for each client, person and project that describes where things stand today. That note is rewritten whenever something changes, and each statement in it links back to the meeting it came from. If you ever wonder where a claim came from, the source is one click away.
Let Software Do the Housekeeping
Any knowledge base falls apart once it relies on people remembering to update it.
So after each meeting, an AI agent writes the meeting up, pulls out the decisions and updates the notes for the people and clients involved. Software then rebuilds everything that can be worked out from those notes, such as each client's timeline and decision log, and checks that nothing is missing or inconsistent.
We also set some firm limits. The agent is not allowed to guess. If nobody stated a person's job title, for example, it leaves the title blank.
Leave Live Numbers Where They Live
We don't copy anything that changes from day to day, such as deal values, tasks, hours or bank balances. A copied number is out of date almost immediately.
The note for a client simply points to their record in our pipeline. If you ask what a deal is worth, the AI looks it up there and gives you today's figure.
What Our Company Brain Runs On
Nothing here is exotic, which is the point.
- The notes are plain text files.
- Meetings are recorded and transcribed automatically.
- The AI agents run on large language models. We use more than one, and they all work from the same written instructions that tell them where each kind of information lives.
- Everything live stays in its own system: tasks in our task tracker, money in our accounts, documents in shared storage, and deals in our pipeline.
A larger business would more likely keep the same design in a graph database. Where it is stored matters much less than getting the structure right.
How the AI Answers a Question
Every note starts with a one-line description of what is in it, so the AI doesn't have to read much to find its way. It checks an index, opens the one or two notes that look relevant, and follows their links if it needs more. It only goes back to a full transcript when nothing else has the answer.
The video at the top of this page shows this happening on a copy of our own notes, with every client and person replaced by a fictional one. These are some of the questions we put to it:
- "Have we migrated Looker to Omni before?" It finds four past projects in the index and opens the closest match for the detail.
- "What did we decide with this client?" It reads that client's decision log, a single page of about 5KB. A person would have had to go back through eleven meetings.
- "Who do we know at this company?" The client's note already lists everyone we have met there and how many meetings each of them has been in.
- "What is this deal worth today?" It doesn't try to answer from the notes. It passes the question to the pipeline, which holds the live figure.
Each answer comes with numbered references to the notes it used, so you can check it yourself. You can try the same questions in our interactive explainer of how the company brain works.
The Results
The copy in the demo holds 1,050 notes with 16,619 links between them, and 729 decisions pulled out of meetings into a log for each client. We counted those in November 2026.
Every question in the video is answered from one or two notes, and none of them needed a transcript. The AI read between 1KB and 22KB of text per question.
For us, the bigger change is that a question about a client no longer depends on who was in the meeting. Anyone on the team can ask, and so can any of our AI agents, and they all get the same answer from the same sources.
Those agents also work across our CRM, time tracking and accounts, and a person approves anything that changes a record. Our AI chief of staff case study covers that side of it.
Company Brain FAQs
What is a company brain?
A company brain is a single, structured store of what a business knows, kept so that both people and AI agents can use it. It holds the things no other system does: who people are, what was said, what was decided and why.
How is a company brain different from a knowledge base or wiki?
A wiki is written for people to browse, and it depends on people to keep it current. A company brain is built for AI to read as well. Every note describes itself, everything is linked, and software maintains the lists and logs, so it stays accurate without anyone tidying it.
Is a company brain the same as RAG?
No. A typical RAG setup searches raw documents and hands the AI whichever passages look most like the question. A company brain gives the AI structure first: an index, a current summary and links to the source. In our demo, that meant every question was answered from one or two notes.
Do you need a graph database to build a company brain?
No. Ours is plain text files, and that has been enough to run the business on. A graph database makes more sense once you have many teams, far more records, or other software that needs to query the brain directly.
Can the AI change things on its own?
It writes up meetings and keeps the notes current. Anything that changes a record in our CRM, ledger or accounts is shown to a person for approval first, and anything going to someone outside the business is only ever a draft.
What We Would Tell Another Business
The tools matter less than you might expect.
What made this work was deciding where each kind of information lives, writing the rules down where both people and AI can read them, and letting software maintain anything it can work out for itself. None of that depends on the size of the business or the technology it already has.
Want a Company Brain of Your Own?
We turn how your business thinks and operates into a brain AI can use.
If you would like to see ours answering real questions, or talk through what one would look like for your team, get in touch.
