Your judgement is unprotected. Own the layer it lives on.
Intelligence is free. Judgement is the moat. Every company will become an intelligent organisation, running on the same models as its competitors. What sets it apart is the judgement it brings to them, and that judgement sits in a few heads, leaves when they do, and leaks into someone else's model.
Private inference protects your data. Nothing protects your judgement. The protection is ownership of the layer it lives on. We help a company build that layer and operate it. The company owns it.
Where your judgement lives, and what holds it.
The layer sits between your people and any model. What goes into it differs for every company, so the extractors that feed it do too.
Where judgement leaves a trace (illustrative)Every company leaves a trace of its judgement somewhere different.
Where judgement leaves a trace (illustrative)Often in client conversations and in how documents get revised.
Where judgement leaves a trace (illustrative)Often in code review, design decisions and incident threads.
Where judgement leaves a trace (illustrative)Often in site decisions, change orders and who got the call.
- Email, chat, meetingsExtractors in the Lab
- Documents and contractsPartly in the Lab
- Code, tickets, runbooksBuilt when a client needs it
- Site and field recordsBuilt when a client needs it
- ERP, CRM, structured dataWith a partner, if needed
Running in the Lab Built for you when you need it
- EvidenceWhat was said, with the source text.
- MemoryWhat the company has decided and learned.
- PermissionsWho may see and change what.
- DecisionsOwner, outcome and what followed.
Runs where you choose: inside your network or outside it. Your call.
- Send only what is neededA task gets the context it requires, not the whole archive.
- Check permissionsWho may ask, and what they may see, is enforced here.
- Record every callWhat left, which model, and what came back, on a log you own.
- Verify what returnsAnswers are checked against the source before anyone relies on them.
- Hosted frontier
- Open, in your network
- Specialist
- BriefsBefore a decision, the evidence and the history in front of the person deciding.
- AnswersAsk the company a question. Get the answer with its source.
- AgentsRoutine work done within the permissions you set.
- SignalsA flag when a decision drifts from what the company has learned.
- ProductsA solved problem, packaged so it repeats.
Every outcome is recorded back into the layer, and the next decision starts from it.
Built and operated with you by IntellumiaYou own the layer
What we believe, and why.
Intelligence is free. Frontier models supply, at close to zero cost, most of what a company once paid people and advisers to know. Two firms on the same models converge on the same decisions, so intelligence alone no longer separates them.
Judgement is the moat, and nothing protects it. Private inference protects your data. It does not protect your judgement, and the people who build these systems say they do not fully understand them (Amodei, April 2025; OpenAI's Private Intelligence, 29 September 2026). A clause is not control.
Own the layer. Rent the model. The layer is a company's own system of evidence, memory, permissions and decisions, owned and controlled by the company, on any model it chooses. Organisations forget. Models do not. The layer is how a company remembers.
One problem at a time, until it is solved.
We start inside a company, on one problem that is specific to it, with a number agreed before we begin. We stay until it is solved. Where you are today is where the work starts, which is why we embed with your team rather than arrive with a product.
When a solution proves it repeats, it becomes a product, on its own or combined with another. That is how a platform gets built: from solved problems, not from a roadmap. The platform is where this ends. The method is how we get there.
The company keeps its layer throughout. We bring the know-how to build it and operate it. Where it runs is your decision. Inside your own network is our first suggestion, with the trade-offs laid out, and you make the call.
Not the answer to everything.
We do not serve every company. We serve every unique problem, until it is solved. We are not the answer to every problem, and we are not experts in every problem. We will never tell you that only Intellumia can solve it. Your judgement is your moat, never ours. What we build with you belongs to you.
We would revise this if we found any of these.
We will review the thesis against our evidence in April 2027.
- Most of a company's judgement cannot be made explicit enough to improve a decision.
- Model makers give customers durable memory that they own and can take to another model.
- A generic model, given no company history, decides as well as the company's best people.
- Buyers still hand their decisions to a single vendor once the trust argument is understood.
Read the full argument, with sources (Thesis v1.0, October 2026).