Most small companies run on memory. Not documents, not systems: memory. What the company knows lives in the founder's head, in two or three senior heads, and in ten thousand messages of chat scroll that nobody will ever read again.
It works, right up until it doesn't. Someone leaves and takes a department's worth of context with them. A new starter takes months to become useful because the real manual was never written. The same decision gets made three times in a year because nobody recorded why it was made the first time.
I run a CGI studio and a search consultancy. Both had this disease. Writing was the cure, and it turned out to be worth far more than the sum of the documents.
Oral tradition is expensive
The costs never appear as a line on the accounts, which is why they survive. The explanation repeated for the tenth time. The meeting that exists because the answer lives in someone's head and that someone is busy. The work redone because the standard was never stated. The proposal that starts from a blank page when most of its content is the same every time. The key person who cannot take a holiday without the company slowing down.
Add it up honestly and most small companies are paying a salary's worth of forgetting every year.
The same idea with harder rules
I wrote separately about building a personal second brain. The business version is the same plain idea, one written home for what the company knows, with three differences that change everything.
It has to be agreed, not just written. A private note can be half-finished and half-wrong. Shared canon has to be true, because the first time someone follows a page that is out of date, they stop trusting all of it, and the whole thing dies quietly.
It has audiences. Not everyone should see everything. Access follows role, and some information does not belong in the shared brain at all.
And it has to outlive people. The point is that what the company knows stops depending on who happens to be in the room.
Decisions are the highest-value pages
If I could keep only one habit from all of this, it would be the decision record. What we decided, when, and why, written once, findable by the team.
Re-litigating settled decisions is one of the most expensive habits a company can have, and it is invisible because it looks like healthy debate. It is not. It is the same meeting on a loop. A decision that was never written down will be made again, and there is no guarantee it will be made the same way twice.
What AI changes for a team
This is where the writing pays back at a different scale. With a shared brain, AI stops being one person's productivity trick and becomes company infrastructure. Drafts start from your actual services, your standards and your voice instead of a blank page. Briefs, proposals and reports assemble from what is true. New tools become useful on day one, because the context they need is already written down.
The rules I hold personally get held harder here, because the blast radius is bigger. Assistants draft; people approve. Nothing reaches a client or the public without a human deciding it should. Client work stays out of tools that remember it, and every automated step is tested before it is trusted.
Boundaries make it trustworthy
This is the part most teams skip. A shared brain without boundaries becomes one of two things: a dumping ground nobody trusts, or a leak waiting for a forwarded link.
Some knowledge belongs to the whole team. Some belongs with the founders. Some, like client commercials and anything personal, needs a smaller room again. Deciding those boundaries early, and holding them, is what makes the rest of the system safe to rely on. People contribute honestly to a system they trust, and clients extend more trust to a company that visibly handles information like adults.
Start with one page
The temptation is to buy software and announce a knowledge base. Resist it. Start with one page: how we do the work here. Then let meetings end with decisions written where the team can find them, not minutes nobody reads. The habit is the system. The software is furniture.
Add AI once the writing exists, not before. An assistant pointed at an empty knowledge base automates nothing but the blank page.
I have argued before that the founder's job is standards: holding a line until it becomes other people's habit. A business second brain is where that line gets written down. A company that records what it knows stops depending on who is in the room. That is the whole point of building one.