Company brains are a hot topic in Silicon Valley right now. AI is making them the new base layer of everything. We’ve built one internally at my new startup, Rustproof, and we are helping clients create them as well.
A company brain is a stored knowledge base made up of meeting transcripts, emails, slack and teams messages, your shared and personal document drive, etc. Everything is ingested and stored so it’s searchable. Then you build AI agents and skills on top of it to use the data in your workflows.
Perfect is the enemy of the good
But recently a client told me he didn't want one for his organization or at least not one based on the data that already existed. The data might be bad. Old contracts, wrong numbers, half-finished projects, notes that contradict each other. Why pour all that into a system and then trust what comes out?
Fair worry. Garbage in, garbage out. This is the oldest truism in computing.
I let it slide. We’d do it piece by verified piece.
But thinking about it later, I was kicking myself. This was the wrong approach. First of all, that’s going to take forever and require huge lifts from the most time-constrained people in the organization.
Second, I doubt anyone would use a system based on that verified data because it wouldn’t be true. Because you're already running on the “bad” data. It exists because people need it to do their jobs.
Every decision your team makes comes out of that same pile: the stale contracts, the wrong numbers, the Slack thread nobody can find. The mess isn't a reason to keep the data in the dark. The mess is what you've been managing blind.
The question that occurred to me is if you can't bear to look at your own data, what does that say about how you run the company right now?
Your company brain already exists
It runs every day in your teams, your email, your documents, your meetings. The map of who knows what, which customers actually matter, why the last big deal died, where the money leaks. That map isn't sitting in a folder waiting to be used. It's already running your company, and it has been the whole time.
That map is called a knowledge graph . That’s a fancy term for what you call things in a company and how those things relate to each other.
What does a client mean and what does a project related to that client mean and how are they connected to invoicing and your services?
For example, if you call the marketing department of company a client but the larger company is an account, that’s part of your graph. Is a project part of an engagement or is a project what other companies call engagements? Who do you actually bill for the work? The marketing department or procurement at the company? What do you call them?
That’s your knowledge graph and there’s one already operating inside every organization. You just can't see it.
Facebook had a similar concept when it talked about your friend graph. You already had friends. Facebook just documented the connections in software.
Why a knowledge graph is valuable
When you document your knowledge graph and then encode it in a database, you can start building layers on top of it. Now you know how things connect: this customer, to that deal, to this complaint, to that unpaid invoice. A map of your business, finally written down.
This is what my client and I missed in the meeting. Ingesting all your data is essential to creating your graph but creating it is only step one.
Creating it is what lets you change it. You can't fix a sales process you can't see. You can't cut waste you can't find. Make the brain legible, get it documented, and for the first time you can rewire it on purpose.
This is why most business process changes fail. They can’t see the existing process because it’s subliminal. People will tell you the process and then when you run it you realize they forgot to mention 25 others things that happen or impact the process that they didn’t list.
I had a similar issue with a non profit I’m helping build a brain. They run everything on Google Sheets. When you ask about a process, they give you a sheet with 17 tabs. I ingest it but it doesn’t produce the outcome exactly so I ask why.
“Oh, there’s this other sheet that does that.” We’re on round 5 of that and counting.
The graph compounds
When we build the graph, the AI will surface one of those wrong numbers my client was afraid of. Now he can correct it and the correction gets encoded in the knowledge graph database. The brain stores what you know and what you've fixed. Do that for a year and you own something no competitor can buy, copy, or hire away.
Think about the questions you can't answer today. Why aren't leads converting? What actually happened to that marketing spend? Where are we bleeding money? The answers exist. They're in your email, your CRM, your people's heads. You just can't reach them. The brain is how you reach them, and then how you fix what they tell you.
Understand what you're signing up for, though. This is less a software project than a habit change. Your people have spent whole careers keeping what they know in their heads. It's how they’ve stayed valuable. Asking them to write it down feels like handing over their leverage. Expect resistance. Lead through it anyway.
What’s next?
Three steps. Do them in order.
One — for you. Get Claude Cowork. It’s $20 bucks! Connect your email, your documents, your Slack. Then start asking questions across all of it: "What did we promise this client?" "Pull every pricing complaint this quarter." Anthropic even ships an enterprise search skill that does exactly this. Start there. This week.
Two — for everyone. Once it works for you, get your team doing it. The brain only compounds when the whole company feeds it and uses it.
Three — build the databases. Structured tables, real pipelines, the durable layer underneath. This is the heavy lift and it's the step most people start with, then stall on. Don't. Earn it by doing one and two first and then hire someone to scale it for you.
Your competitors are sitting on the same mess you are, just as afraid to look at it. The edge doesn't go to whoever has the cleanest data. It goes to whoever makes their brain legible first — and starts rewriting it.
— Alan
P.S. Here’s an example of the compounding I mentioned above in action from another great newsletter, Random Walk.
It shows that the top 1% of AI users are using 10x more AI than the top 10%. That is inequality for you. Mostly these are coders but it shows you the kind of compounding people and organizations experience once the building blocks of AI are in place. A company brain is one of them:
“Here’s another cut of a similar observation:
Per Ramp’s spending data (so take it for what it is), the top 1% of token spend/employee is more than 10x the top 10%. Once again, the gap is big and growing.
All of which is to say there’s more fuel to the “AI Power User” fire.
The tech is real and transformative, but a small number of users appear to be way out in front, and they are compounding their lead. And that’s the thing about compounding exponentials: small differences become very big differences very quickly.
And therein lies some of the trickiness of finding “AI winners,” or the broader question of observing “AI ROI” in the 10,000ft picture. (1) If gains are concentrated then, by definition, diffuse benefits will be hard to observe (at first), widespread margin growth, notwithstanding; and (2) if you thought we had pareto distributions before, well, you ain’t seen nothing yet.”