A brain for every employee. One that ties them together.

Most companies installing AI make the same architectural mistake. They either centralize everything — one shared workspace, all knowledge funneled into the corporate vault — or decentralize everything: everyone gets their own ChatGPT subscription and figures it out. Both break within twelve months.

• Centralize-everything fails the employees.The knowledge people build over years lives in their heads, their email, their habits — and most of it can’t move into a corporate AI without losing the context that made it useful. People resist contributing because what they upload becomes the company’s property the moment they upload it, and they intuit that this costs them leverage. The shared brain ends up half-empty.
• Decentralize-everything fails the company.Each person builds their own ecosystem of prompts and projects. Institutional knowledge that should compound across the team doesn’t — there’s no shared layer. Two people on the same problem work with a different AI and a different mental model. The aggregate effort is wasted.

The Company Install resolves this by being explicitly two-tier. Every employee gets their own Personal Brain — owned by them, portable when they leave, fed by their own work. The company gets a shared Company Brain — owned by the company, fed by what it has decided is institutional knowledge. The two query each other through a governance layer that respects the line.

Two parallel brains, joined by a governance layer.

• Each Personal Brain holds the human’s own context. Same architecture as the solo Install— a private, gated brain queryable from any AI tool — but what’s inside it is the person’s: the books and notes they collect, their takeaways, the judgment and skills they’re building. That human layer is what makes an AI’s output land right, and what develops the person. It’s theirs, and it makes their work better.
• The Company Brain holds the work product — and all the knowledge around it.Institutional knowledge (policies, playbooks, frameworks, case studies) and the workflows and skills people build and contribute — but also any database, tool, or system you give us access to. Anything with an API, we wrap into an MCP server the brain can query (often a CLI, MCP, or API already exists), so we can actually pull your data in. Done right we ingest it in bulk up front rather than a slow trickle, and chunk it into the brains it belongs to. It’s owned by the company, on company SSO (M365, Google Workspace, or Okta). It stays when people come and go.
• A “Company Brain” is usually a cluster, not one brain. What shape it takes depends on the data patterns the Working Intelligence Audit surfaces — and we wire it accordingly during the install:
  • By department, or sub-department — an HR brain might split into a talent-strategy brain under the direct HR lead, where the work is genuinely distinct.
  • By data shape — a marketing brain might be a cluster: one for media, one for text and prose, one for the rest, because each is queryable best in a different shape.
  • Or a single brain — learning & development, say, where one is enough.
  Either way they can be stitched into one tool an AI queries as a whole, or wired in separately — the audit decides, and the install does the wiring.
• The governance layer is the connective tissue. Most installs use a three-tier model:
  • Personal → Personal — off by default; cross-brain rights only open when both employees opt in.
  • Personal → Company — on by default, scoped by permissions.
  • Company → Personal — off entirely; the company never automatically reaches into individual brains.
• The endgame is orchestration.Once the company’s knowledge actually lives in the brain — not just documents, but the data, the tools, the systems — it stops being a place you look things up and starts running the work: an orchestrator coordinating agents, workflows, and routines at a granular level. That’s roadmap, not week one. You start from a strong foundation to get there quickly, and for real.

Every skill installed comes with a correctness contract — how the agent proves it got the right answer, what it’s allowed to claim, and how you reverse a bad call.

You can’t have a Company Brain without people who’ll actually use it through their own.

• Shared-brain-only installs become write-only archives.Deploy just the shared brain and employees keep using their personal ChatGPT and Claude, because that’s where their workflow already is. The investment gets shelved within a year.
• The fix is to deploy from the user out.Each employee’s Personal Brain gives them an immediate reason to engage. The Company Brain then earns its keep because they’re already querying through tools that reach it. The shared brain compounds because the personal brains pull from it constantly.
• Three is the floor— below that, the cross-brain governance design isn’t worth the complexity. At three, the shared layer starts paying for itself. At ten, it’s indispensable.