One brain for everything your company knows
Your scattered knowledge compiled into one structured, searchable source that staff and every AI tool you run can query.
The problem
How it works by hand
Every small business has its knowledge scattered across inboxes, shared drives, old proposals and the head of one long-serving employee. New starters take months to absorb it, answers depend on who happens to be in the office, and when that one person is on holiday the business gets measurably worse at its own job. Meanwhile every AI tool you try gives generic answers, because it knows nothing about your business.
A worked example
What a working version looks like
We start by gathering the raw material: documents, proposals, case notes, email threads, pricing history, the folklore that only exists in conversation. AI compiles it into a structured knowledge base, organised the way your business actually thinks: by customer, by service, by sector, whatever the natural shape is, with every entry traceable back to its source. That structure is then exposed two ways. Staff query it directly through an assistant that answers in plain English and cites where the answer came from. And every other AI system you run, from proposal drafting to customer service, plugs into the same brain, so it writes and answers with your real history rather than guessing.
The exact tools change per business. The shape does not.
What it needs
Honest inputs, nothing exotic
- 01Access to where the knowledge lives (drives, inboxes, old proposals, exports)
- 02A few hours with the people who hold the unwritten version
- 03A decision on who can see what, if some material is sensitive
- 04Somewhere to host it, or we host it for you
The payoff
What you get back
Answers stop depending on who is in the office. New starters get productive faster, hard-won knowledge survives people leaving, and every AI system you build afterwards gets dramatically better, because for the first time the AI genuinely knows your business.
Do it yourself
How you would build this yourself
No course, no upsell. This is the order we would build it in, with the tools named, and a prompt to start from.
- 1
Gather the raw material before buying any tool: proposals, case notes, pricing history, the email threads where decisions actually got made.
- 2
Interview the people who hold the unwritten version. Two hours of recorded questions with your longest-serving person beats a thousand documents, because the valuable knowledge is in heads, not drives.
- 3
Structure beats search. Have AI compile the material into organised entries (by customer, by service, by sector) with every claim traced to its source, rather than dumping files into a vector database and hoping.
- 4
Decide access before launch: who sees margins, salaries, sensitive clients. Retro-fitting permissions onto a knowledge base is miserable.
- 5
Expose it two ways: a chat assistant for staff that cites where each answer came from, and a format other AI tools can read, so everything you build afterwards gets smarter.
- 6
The fiddly bit is keeping it current. Give it one named owner and a monthly review, or it rots like every company wiki before it.
Help me build a structured knowledge base for my business. The sources are in [folders/drives/exports I will share]. Work through: 1) inventory and read the material, 2) propose a structure that matches how the business actually thinks (by customer, service, sector), 3) compile entries with every fact traced back to its source document, 4) flag the gaps where the knowledge is probably in someone’s head, with the exact questions I should ask them, 5) set it up so I can query it in plain English and other tools can read it. Start with the inventory.
Copy it into Claude Code, fill the brackets, and it will plan the build with you before writing a line of code.
We would rather show you how than bill you. The whole ladder of free help, answers, guides and the weekly build-along, is on the do-it-yourself page.