Connected is not the same as understood.

Each of your systems holds a piece. The CRM knows the customer, billing knows the invoice, the inbox knows the complaint. On their own they are records, and a question that crosses them gets answered by whoever has the patience to look in all three.

The brain connects them into one picture: this invoice belongs to that customer, who raised that complaint, which is why the renewal is at risk. That web of connections is a knowledge graph, and it is what lets a question be answered rather than searched for.

The links are authored, not guessed: by people, or as each note is written. The brain does not build its graph by having AI read everything and decide what relates to what.

Second brains, personal knowledge graphs and notes apps with AI on top all work for the same reason: one person owns them. They write everything in, they may read all of it, and when they leave, it goes with them. Nobody else has to rely on it.

A company is the opposite on every count. Hundreds of people read from it, and most must not see everything. Many write to it, so someone has to decide what counts as true. It holds personal data, so the law has a say in what goes in, who sees it and when it must come out. And it has to outlive everyone who built it.

That is why a company brain is built the way it is: an owner and a source on every fact, permissions enforced by the system, and a record that stays with the company. The sections below take each in turn.

The things your business runs on, and how they relate: customers and contracts, decisions and the reasons behind them, policies, people and roles, commitments, and the numbers along with what each one is defined to mean.

And the part no system holds on its own: what your company means by a good outcome, which exceptions matter and which are noise, how decisions actually get made as opposed to how the process document says they do. That is the part no vendor can sell you, because none of it is generic.

A shared drive stores everything: drafts, duplicates, the version that was withdrawn in March. A search over it finds all of them, and an AI tool reading it cannot tell which one counts.

The brain holds what someone has decided is true, with an owner and a source on every entry. A named person approves every source before anything is captured, so nothing goes in by accident.

You decide, and the brain enforces it. Who may see what is built into the system, not into a polite instruction the AI is asked to follow, and your rules decide what AI and agents can see and do.

It is built to use your own sign-in, so when someone leaves, their access goes with their account, and every permission expires anyway unless someone renews it. Every question, answered or refused, is recorded with who asked it and which sources the answer drew on.

Some things should never go in at all. We agree that list with you before anything is captured, not after. When someone’s personal data has to come out, the brain is built to erase it from the record and from everything built from it. We have rehearsed that on our own.

When AI reads or answers from it, the passages it needs go to that model’s provider, through your own business account, not ours. For your most sensitive material, we can build it on a private or open-source model that runs in your own environment.

Every answer carries its source, so anyone can see where it came from. When the evidence is not there, the brain is built to say so, and why, rather than guess.

That matters more than it sounds. A model asked to fill a gap in what it knows will fill it fluently, and wrongly; a model given the context retrieves instead of inventing. Why grounding matters more than the model.

A brain nobody curates goes stale, so we agree who keeps it current before we build. If we hand it over, your team does, with what it needs to do it. If we run it for you, we do: new material taken in under your rules, every source reviewed on its own date, mistakes corrected at the source. Two ways to have it →

Either way it is yours: plain files, an index that rebuilds from them, on your own servers or in your own cloud account, and nothing locked into any vendor, us included.

The clearest way to see what understanding your company would take is the question audit: about two hundred real questions from your business, answered in writing with what a company brain would answer well and what it would not.

The thinking behind it is in A company your model can read, and you can run a first version yourself with the context audit skill, in your own AI assistant.

Start with a conversation. Book a discovery call, or try the ten-minute diagnostic (opens in a new tab) yourself.