Brain
What Brain is
Your team's knowledge as a linked set of pages — people, companies, concepts and sources — written by both your team and the AI, and used to ground every answer about your organisation.
Brain is what your organisation knows, written down in a form the assistant can use. Policies, processes, people, customers, decisions, the reasons behind them.
It exists to answer one question: when someone asks the assistant about your company, does the answer come from your team, or from what a model assumes companies usually do?
It is a wiki with two writers
Everything in Brain is a page — a title, some markdown, and links to other pages. What makes it different from a wiki is who writes it: your team types pages, and the AI proposes pages from material you give it. Both go through the same rules, the same history, the same review.
Pages link with [[double brackets]], and those links form a graph. Linking to a
page that does not exist yet creates a stub — an empty page standing where
knowledge should be. Stubs are not a defect; they are the map of what you have
not written yet.
Pages have types
Every page is a kind of thing: a person, a company, a concept, a note or a source. Your workspace can add its own categories — Meeting, Product, Client — and switch off ones you do not use. Types are what let the brain answer “who is this?” differently from “what is our policy on this?”.
See Brain settings for how categories are managed.
Four things you do with it
| Move | Page |
|---|---|
| Teach it — paste, upload, link, import | Teaching the Brain |
| Review what it proposes before it lands | Reviewing what it proposes |
| Browse and edit the pages themselves | Browsing and editing pages |
| Ask it questions and see the gaps | Asking the Brain |
Where it gets used
Brain is not only a page you visit. It is read wherever answers are produced:
- In any task, through the company knowledge tool — the assistant searches it before answering, and cites the pages it used
- By AI staff connected to it, so unattended work is grounded in your material rather than the model’s assumptions
- By your own apps, for signed-in members
Why it is worth the effort
An ungrounded model asked about your parental leave policy will produce an entitlement, a notice period and a clause number. All fluent, all invented, all indistinguishable from a real answer.
With a brain, the same question returns your policy, the page it came from, and — if you never wrote one — the plain statement that nothing covers it. Both of those outcomes are useful. The invented one is not.
Where it lives
Brain belongs to the workspace. One brain per workspace, visible to its members, with owners and admins deciding how it is configured and what gets approved.