The Company Brain
Alfred taught me that a person with a connected knowledge system outperforms a person with a pile of saved links. The company brain is the same idea at business scale: one Notion workspace that knows how the whole operation works, readable by every employee and by the AI they use, built from nothing inside a company that had no documentation at all.
One source of truth
The brain holds the company itself in writing: every process, documented; every person, what they do and what their job covers; hundreds of supplier documents, guides, and terms digitised out of drawers and inboxes. Core operations and inventory workflows were raised to versioned procedures with named owners and escalation paths. When someone asks how something works here, the answer is a page, not a person's memory.
Wired into Claude
The part that makes it a brain rather than a wiki: the whole team's Claude accounts are connected to it. When anyone in the company asks their AI a question, the answer is grounded in our actual processes, our roles, and our data, not in generic knowledge. New starters onboard from it, the AI reasons from it, and my agents draw their context from it. One brain, many readers, and all of them on the same page.
The AI fast path
Underneath sits a two tier memory system for the AI itself. A compact working memory loads into every session: who everyone is, the company glossary, live project state, and a log of hard won platform quirks. Behind it sits a library of deep context documents pulled in on demand. Its most valuable job is decoding the company's internal shorthand, the grades, the trade terms, the nicknames, so an agent parses a request the way a colleague would. The hierarchy is explicit: Notion is the human source of truth, memory is the AI fast path, refreshed when the truth changes. Every agent in the fleet is shorter, sharper, and less supervised because this layer exists.
The context pack agent
Once heads of department had their own Claude accounts, they each needed the same context. So the brain maintains itself: a weekly agent scans the key pages for material change, and only material change, a project opened or closed, a person or system changed, a policy updated. If something real moved, it updates the affected sections of a versioned context pack, writes a changelog line, and nudges the team on Slack to refresh. If nothing moved, it stays silent. And if it cannot be certain it is posting to the right channel, it drafts the message for a human instead of broadcasting. Restraint, designed in.
The rollout is the hard part
A knowledge system nobody uses is a graveyard with better formatting. This one shipped as part of the company's first integrated tech stack, on a one month migration plan, with CEO sign off and training for every head of department, run live in the room rather than sent as instructions. Adoption was designed, not hoped for. That is the real lesson of the project: the writing is cheap, the trust is the build.