---
title: "How to build a company brain on your data warehouse | Rig"
description: "A company brain is a governed map of your company data that staff and AI agents can ask, built on the warehouse you already run. Here is the build order."
canonical: "https://rig.so/blog/how-to-build-a-company-brain"
format: markdown
---

[Data context and governance](https://rig.so/blog/data-context)

# Can we stop calling a chatbot on raw tables a company brain?

A company brain is your company data in warehouse tables, plus a map of what those tables mean, which staff and AI agents can ask through a governed gateway. It is another name for a context layer on the warehouse you already run. A chatbot pointed at raw tables can give two people two different numbers for the same question.

By [Toby Mather](https://rig.so/blog/authors/toby-mather) · Published 20 November 2026

[Book a call with an architect](https://rig.so/book-demo?offer=architect_call&from=%2Fblog%2Fhow-to-build-a-company-brain)

## Key takeaways

- Company numbers belong in warehouse tables, where the same question returns the same answer.
- The brain is the map over those tables: definitions, joins, certified metrics and usage rules.
- A gateway decides what each person and agent can read, and logs every call.
- Build it one team at a time, starting with work people already do every week.
- A company brain holds numbers in tables, so handbooks and policies can stay in your document store.

1. 01**CRM, billing, product events**
   
   The systems the business runs on
2. 02**Warehouse**
   
   Customers, deals, tickets and invoices as tables
3. 03**Map, metrics, relationships**
   
   What each table means and which metric is certified
4. 04**Gateway**
   
   People and agents, each with their own role

Sources land in the warehouse, the map explains them, and people and agents ask through one gateway.

## What do we mean by a company brain?

People use the phrase for anything from a shared drive with search to a chat window over the CRM. We use it for one specific thing: the company’s operational data, held as tables in a warehouse, with a map that tells people and AI what each table means, how the tables join and which definitions are certified. Staff and agents ask questions through a gateway that applies their permissions.

That makes it a plain-English name for a context layer, and there is no separate product called a company brain. The phrase here has nothing to do with brain.co or brain.ai, which are separate companies.

## Why is a chatbot on raw tables a bad idea?

A model pointed at raw tables sees column names and guesses the rest. It finds three columns that could be revenue and two customer tables, and picks one each time it answers. Ask for revenue by segment on Monday and again on Tuesday, and you can get two different numbers with equal confidence.

It also tends to run on a shared login, so it can read everything that login can, and the logs cannot tell which person asked. Neither problem shows up in a demo, which is why a pilot can look good until finance checks the figures.

## Why do notes and documents make a poor store for numbers?

Teams build AI brains from markdown files, Notion pages and document search because they are quick to write. They suit handbooks, policies and onboarding guides. Numbers copied into them go out of date: a note that says “the biggest customer is Acme” is correct until next quarter, and nothing updates it when the CRM changes.

Facts about customers, deals and invoices belong in rows and columns in the warehouse, synced from the systems that own them. The map over those tables records what they mean, and it can be kept current because it is checked against the schema.

## How do you build a company brain?

Use the same build order as an enterprise context layer, one team at a time. Start with work people already do every week, such as checking invoices, preparing a trading report, reviewing accounts or investigating a failed payment. The business owner defines the task and how to measure the result, and the data team controls the definitions and access.

1. Bring the data in: sync the team’s tools into a warehouse, or use the warehouse you already have.
2. Map it: describe the tables and columns, detect joins and draft the metrics the team uses.
3. Write down the rules: record the definitions and exceptions the team already works to, such as which deals need CFO sign-off.
4. Certify the metrics: compare each one with a figure its owner already trusts, then have that owner certify it.
5. Connect the tools: give people and agents access through a gateway, each with their own role.

## Why does a company brain need a gateway?

Once Claude, Cursor, Slack bots and scheduled agents can all ask questions, access has to be decided per caller. A gateway gives each person their usual permissions and gives each agent its own role, token and call limit. Row rules mean a sales rep sees their own deals and a manager sees the team, in a chat window just as in a dashboard. Every call is logged by name, so when a number is questioned you can see the metric and the query behind it.

## What does it look like in practice?

At Birdie, RevOps replaced fixed legacy BI dashboards with certified metrics on Snowflake, and finance, customer success and sales built on the same data. Birdie’s customer success team now trusts the answers it gets from AI. At Cleo, one Growth Specialist’s agents made 8,023 requests to company data in 30 days, on certified metrics that Cleo’s data team reviews.

In both cases one team set up and reviewed the certified metrics, and other teams then asked questions against the same definitions. To repeat it, start with one team’s weekly work and get those numbers right before you add the next team.

## How does Rig do it?

Ingest syncs your tools into a warehouse you own, or Rig works on your existing Snowflake, BigQuery, Databricks, Fabric or Postgres warehouse. Rig Map drafts table descriptions, joins and metrics from the warehouse, your team reviews and certifies them, and usage rules record the exceptions. Rig Map updates as schemas change.

Gateway connects Claude, Cursor, Slack and your own agents through one MCP connection, with each person’s or agent’s permissions applied and every call in the audit trail. Answers show the metric and the query behind them, so anyone can check where a number came from.

## Common questions

Is a company brain a chatbot?

A chat window is one way to ask it questions. The brain is the governed data and the map underneath, which Claude, Slack, dashboards and agents all read.

Do I need a new warehouse?

No. A company brain works on the warehouse you already run. If you do not have one, a managed ingest service can sync your tools into a warehouse you own.

How is this different from a knowledge base?

A knowledge base holds documents such as handbooks and policies. A company brain holds the numbers, as tables read with SQL, with a map that tells AI what each table means.

Is this a different product from a context layer?

No. Company brain is a plain-English name for a context layer on your warehouse, with a gateway in front.

Which team should go first?

Pick a team with a weekly task that depends on numbers, such as a trading report or invoice checks, and an owner who can say whether the answers are right.

## Related reading

[Data context and governance](https://rig.so/use-cases/data-context-and-governance)[Company brain](https://rig.so/use-cases/company-brain)[Rig Map](https://rig.so/products/rig-map)[Gateway](https://rig.so/products/gateway)[Ingest](https://rig.so/products/ingest)[Nobody needs the whole enterprise context layer on day one](https://rig.so/blog/how-to-build-an-enterprise-context-layer)[Everyone's banging on about context, but what actually is it?](https://rig.so/blog/what-actually-is-context)[WTF is a semantic layer anyway?](https://rig.so/blog/wtf-is-a-semantic-layer)[Birdie case study](https://rig.so/case-studies/birdie)[Cleo case study](https://rig.so/case-studies/cleo)

## Sources

Vendor documentation checked on 20 November 2026. Confirm current features and terms with the vendor.

- [Model Context Protocol](https://modelcontextprotocol.io/)
- [Rig case study: Birdie](https://rig.so/case-studies/birdie)
- [Rig case study: Cleo](https://rig.so/case-studies/cleo)

## About the author

[Toby Mather](https://rig.so/blog/authors/toby-mather), Co-founder and CEO, Rig. Toby co-founded Rig after seeing how hard it was for business teams and AI to work with complex internal data. He was previously co-founder and CEO of Lingumi and CPO at Novakid.

Talk to Rig

## Talk it through with an architect

An architect spends 30 minutes with you on your systems, your data and what you want to change.

[Book a call with an architect](https://rig.so/book-demo?offer=architect_call&from=%2Fblog%2Fhow-to-build-a-company-brain)

---

- [This page as HTML](https://rig.so/blog/how-to-build-a-company-brain)
- [Site map for language models](https://rig.so/llms.txt)
- [API and agent documentation](https://rig.so/developers)
