What is a company brain? A plain definition for 5 to 200 person companies

A company brain is a single searchable memory of everything your company knows: every meeting, every customer conversation, every decision and the reasoning behind it, with AI agents on top that do the repeat work. You ask it a question from Slack or a text message, and it answers in seconds with the source attached.

That covers the term as operators use it. Search engines muddle it, so two clarifications before we go further. A company brain has nothing to do with Neuralink or any neurotech startup. There's no hardware and no implant, and if an AI search answer pointed you at a brain-computer company, it grabbed the wrong noun. It also differs from corporate brain drain, which names the problem of experienced people leaving with what they know, where a company brain names a fix for it. If you run a company with 5 to 200 people, the rest of this page defines the term part by part, shows what goes into one, compares it fairly with the tools you've probably tried, and walks through cost and payback with the math shown.

The plain definition, part by part

A company brain is four things working together: a memory, a set of agents, an answer layer and an ownership model. Each part earns its place, and a system missing one of them tends to collapse back into a folder nobody opens.

The memory. Every meeting gets transcribed and filed. Every customer conversation lands in the same place whether it arrived by email, phone or a support thread. Every decision gets written down with its reasoning: what you chose, what you rejected, and why the tradeoff went the way it did. Most companies hold fragments of this in Slack scrollback, inbox threads and a few heads. The memory gathers it into one body of plain files that software can search.

The agents. On top of the memory sit small programs that handle the work your team repeats every week. One drafts the follow-up after every sales call. One sends a briefing before every meeting so nobody walks in cold. One assembles the weekly scoreboard. One chases unpaid invoices. One drafts support replies for a human to approve. Each agent is wired into the tools your team already uses, so adopting the brain requires no change in how anyone works.

The answers. When someone asks the brain a question, the answer arrives with its source: the meeting it came from, the document it lives in, the thread where the call was made. An answer without a source is a guess, and a team stops trusting a system the first time it guesses wrong. Sources are what keep the brain checkable.

The ownership. Everything lives in plain files on machines you control. No vendor database sits in the middle, and no per-seat fee grows alongside your headcount. If the people who built it vanish, the files still open.

What goes into the memory

The memory holds four kinds of records, and together they cover most of what a company forgets.

Meetings. Each one transcribed, summarized and filed where search can reach it. Six months later, when someone asks what was agreed with that vendor in March, the answer takes seconds instead of an archaeology session through calendars and half-remembered threads.

Customer conversations. Sales calls, support threads, onboarding sessions and renewal discussions, all in one place. Patterns show up that no single rep can see from inside their own inbox: the objection that keeps recurring, the feature request that has now come up eleven times, the phrasing that closes.

Decisions with their reasoning. This is the record almost no company keeps, and it is the one that compounds the most. Why you priced the way you did. Why you dropped that product line. Why the refund policy has that one odd exception. Write the reasoning down once and the company stops re-litigating settled questions every time the cast of a meeting changes.

SOPs and process docs. Your SOPs belong in the memory too, next to the conversations and decisions that explain them. Most process docs were accurate the week they were written and have quietly rotted since. Inside the brain they sit beside the live record, so when reality drifts from the doc, the drift is visible and the fix is one edit away.

Why companies build one

Companies build a brain because knowledge keeps leaving and repeat work keeps eating the week. Both problems are measurable, and most operators feel them long before they measure them.

The first failure is concentration. In most 5 to 200 person companies, a handful of people hold the operating knowledge. The ops lead has the pricing history and every vendor quirk in her head. When she leaves, all of it walks out the door with her, and her replacement spends six months re-learning what the company already paid to learn once. Nothing about her exit is unusual; the unusual part is a company that keeps the knowledge when it happens.

The second failure is repetition. The same questions get asked every week, and the same senior people stop what they're doing to answer them. Sales follow-ups slip because writing a good recap takes twenty minutes nobody has at 5pm. Meetings start with ten minutes of reconstruction because nobody prepped. A wiki was supposed to be the source of truth for all of this, and it was, for about a month, until updating it became one more job nobody owned.

Both failures have a price you can estimate. We use one method for it everywhere, so our numbers always agree: hours per week spent on a workflow, times the hourly rate of the person doing it, times 52 weeks. Take sales follow-ups. If two salespeople each spend 3 hours a week writing recap emails and next steps, that is 6 hours a week x $75 an hour x 52 weeks = $23,400 a year on one workflow. Or take the founder who spends 4 hours a week answering questions the brain could answer: 4 hours x $150 an hour x 52 weeks = $31,200 a year. Those lines stack, and in most companies there are more of them than anyone has counted.

A company brain vs the tools you've probably tried

The tools most teams reach for first each solve part of the problem, and each one is right for somebody. The fair comparison looks like this.

Tool	What it does well	Where it falls short for a 5 to 200 person team
Wiki (Notion, Confluence)	A written source of truth for docs and SOPs. Cheap to start, good structure, easy sharing.	Only as current as the last person who updated it. Nothing flows in on its own, and it can't act on what it stores.
ChatGPT Business / Team	A strong general assistant for drafting and research. Per-seat pricing, quick to roll out.	It knows what you upload and only what you upload. No live feeds from your meetings and tools, so its picture of your company goes stale, and it takes no actions in your systems.
Claude Team	A capable assistant with shared projects for team context. Per-seat pricing.	The same gap: you carry context to it by hand, and it keeps no durable, source-linked record of your decisions.
Glean	Enterprise search across the SaaS tools you already use. Finds existing documents fast.	Built and priced for large enterprises. It finds documents that exist; it can't surface decisions that never made it into one, and it does none of the repeat work.
Guru	Verified knowledge cards for support and sales enablement.	A person still writes and verifies every card, so capture stays manual and coverage stays thin.
Company brain (custom build)	Feeds keep the memory current without manual filing. Agents do the repeat work. Answers carry sources. You own the files.	A real build with a real price, and overkill for a team of two or three.

If your team is under 10 people and the main need is drafting help, a ChatGPT Team or Claude Team seat may well be the right purchase, and we'd tell you so on a call. We don't quote their prices here because they change; both use per-seat pricing, and their pages have current figures.

The gap all of these share is the plumbing. You can build a chatbot in an afternoon. The hard part is the feeds that keep the memory current and the agents wired into the tools you already use, and that plumbing is exactly what dies as a side project inside a busy company. A company brain build is, at bottom, paying someone to finish the plumbing and hand you the keys.

What the agents do, and what they do badly

The agents handle work that repeats on a schedule and can be checked against a source. Our standard first set, each one adjusted to the client's tools:

• Sales follow-ups. After every sales call, a drafted recap with next steps is waiting in the rep's outbox, built from the transcript. The rep edits and sends.
• Meeting briefs. Before every external meeting, a short briefing arrives: who you're meeting, the history, open items from last time.
• The weekly scoreboard. The numbers your team reports by hand every Friday, assembled automatically from the systems they live in.
• Invoice chasing. Polite, persistent, and never embarrassed to send the third reminder.
• Support drafts. Replies drafted from the memory of every past ticket, queued for a human to approve.

What agents do badly is judgment. An agent shouldn't decide whether to fire a customer or change your prices. Anything without a source behind it is out of scope, because an agent with no source produces confident guesses, and confident guesses are worse than silence. We scope agents to work where a human can check the output against the record in under a minute, and we'll tell you when a workflow you want automated fails that test.

What one costs and how the payback works

Our builds start at $15,000, and the brain with its first agents goes live in the first month, usually sooner. Before you sign anything you get a fixed quote: one price and a written definition of done. Scope moves the number, and anything that adds build time, like businesses spread across separate workspaces or tools with no standard connection, gets named in the quote rather than discovered later.

After the build there's an optional retainer from $2,500 a month that covers new agents and ongoing tuning. Optional means optional. When the build is paid for, the system belongs to you, and plenty of teams run it themselves from there.

The build carries a 60-day savings review. Sixty days after launch, we run a cost review from the brain's own records. If it finds under $15,000 a year in savings you agree are real, you get up to 2 months of the retainer free.

Payback uses the same math as the examples above: hours per week x hourly rate x 52 weeks, with hours capped so no single workflow claims more than 20% of the team's week. The method prices a founder's hour at $150, engineering at $100, sales at $75 and operations at $55. Run across the top 3 workflows, the audit engine's benchmarks land a 1 to 5 person team around $124,000 a year, a 16 to 50 person team around $300,000, and a 51 to 200 person team around $400,000. Those figures count staff time only, so treat them as a floor; they ignore the deals that stopped slipping and the knowledge that stopped leaving.

If you'd rather see your own numbers than our benchmarks, the free Leverage Audit takes 2 minutes: you give a website, a team size and the areas you care about, and it estimates the 3 workflows where agents would save you the most, with the hours and dollars shown.

The case study we publish

We run our own install, and it carries the numbers we're willing to stand behind. Rye Valley, the holding company behind Hockey Stick, removed $425,000 in annual expenses and raised shipping velocity 10x within the first month after install. The decision log behind those changes is public at michaelhouck.co/decisions, so you can read the reasoning behind each one rather than take a summary on faith. That log is also a working sample of what a decision record looks like inside a brain.

Who owns it and where your data lives

You own all of it. The memory is plain files in formats you can open without us, stored on machines you control. No vendor database sits in the middle, so there's no export to beg for and no migration hostage situation if you ever want to change how it runs. We see what you grant us access to during the build and the retainer, and you can revoke that access any day you choose.

This matters more for a brain than for most software, because the brain ends up holding the most sensitive record your company has: what you decided and why. A system like that should answer to you.

How a build runs

1. An intro call. 30 minutes on your company, your tools and where your team's week goes.
2. A fixed quote. One price and a written definition of done, before you sign anything.
3. The build. The brain and your first agents go live in the first month, usually sooner. From you, we need access to the tools the brain reads and one person who can answer questions during the build. Your team keeps working the way it does now.
4. It keeps growing. We train your team. On the retainer, we add agents and tune the brain every month as your company changes.

A company brain is the difference between a company that remembers and a company that keeps re-buying what it already paid to know. You can start the habit before you buy anything: write this week's decisions down with their reasoning, in whatever tool you have open. That record costs nothing to keep, and it is the seed every brain grows from.