What is a company brain, and how SMEs build one on Slack, Drive and calls
A company brain is an AI system that reads a company’s chats, documents and call transcripts, links people, projects and decisions in a knowledge graph, and answers questions in Slack or Telegram with the original source attached. Small companies can start on ready-made tools in two weeks or build a custom graph in about two months.
What problem does a company brain solve?
UK employees lose about nine hours a week tracking down information, according to Atlassian’s State of Teams 2025. A large share of what a company knows is held by one person: in a Panopto and YouGov survey of 1,001 employees, 42% of company knowledge was unique to the individual. When that person is on holiday or leaves, the knowledge goes with them.
Most of it was never written down. It lives in Slack threads, Telegram groups, call recordings and the shared drive — places a traditional knowledge base doesn’t read.
What is a company brain — and what isn’t it?
It is not a wiki and not a chatbot over a folder. It is a system with three parts:
- Sources: the places where work happens — chat channels, documents, calls, forms — plus the team list as the anchor for people.
- Memory: a knowledge graph of people, projects, clients and decisions and how they changed over time, plus a vector store of the original text.
- Answers: a bot in the messenger your team already uses, replying in a private message or a channel, with the sources shown.
Why use a graph, not just search?
Standard AI search — retrieval-augmented generation over a vector index — finds pieces of text that look similar to the question. On clean documentation that works. On chat history it starts making things up: the same topic is discussed in a dozen places, and half the meaning sits in the relationships between people and projects rather than in any one message.
Ask “who was responsible for the onboarding bot last spring?” and similarity search returns the messages that mention “onboarding bot” and “responsible” most — often about a different bot. A graph follows the relationship instead: bot → led by → Tom, March to May → confirmed by his message of 14 March. The answer is then checked against that original message before it is given. See the side-by-side on our homepage.
What are the ways to build one?
- Ready-made tools. Claude or ChatGPT with native connectors to Slack, email and Drive. Running in about two weeks; good for moderate volumes of clean documents; limited control over retrieval quality and no understanding of relationships.
- A custom knowledge graph. Data collected through APIs, cleaned, entities extracted, a graph plus vector search. About two months; right for noisy chats and questions about people and projects, and where data must stay on your infrastructure.
- A graph with its own interface. The same, plus role-based access and panels per department. A product build for larger companies.
Our prices for each are on the pricing page; the levels are explained on the product page.
How should a small company start?
- Put the data where it can be read. Record and transcribe client calls; move work chat into channels; keep documents in one drive.
- Pick the first questions. List the ten questions people ask each other most. They decide which sources to connect first.
- Start small, measure. One team, two or three sources, a baseline of time spent searching. A client of ours cut search time by 70–80% with over 1,000 messages in the first version.
- Set the rules. Which channels are excluded, who can see what, where the data is stored, and a data processing agreement. See how we handle data.
Not sure your data is ready? The free readiness scorecard takes five minutes.
Questions
Is a company brain the same as Microsoft Copilot or ChatGPT Enterprise?
Those tools can be the first level of one: with connectors to your drive and email they answer from your documents. A custom brain adds the chat history and calls, a graph of who did what, and answers checked against the original message.
Does it work with Microsoft Teams?
Our current version answers in Slack and Telegram. Teams is on the roadmap.
Where is the data stored?
On the supplier’s cloud or on your own AWS, Google Cloud or Azure account, under a data processing agreement. Ask any supplier for this in writing before connecting data.
How we write these guides: figures about our own work come from before-and-after measurements on client projects (names withheld where under NDA); market figures come from the sources listed, with the date we checked them. Prices exclude VAT.