No. A company’s knowledge lives in its organisation - in who decides what, and in how the work actually gets done - not in a data warehouse. Building it is a matter of interviews and sign-off by the people in charge, not an engineering project. A data team remains useful for analytics, but it is not the mandatory gateway to context.
The quote that kills the project
Many AI projects die at the quoting stage. A sixty-person business asks to connect an AI to its internal knowledge, and receives an integrator’s proposal stacking up a data warehouse, ingestion pipelines, a steering committee and a support engagement. A year-long programme, for a question that fitted in one sentence.
Management pushes back, and rightly so. The quote does not answer the question that was asked; it answers a different one, about an infrastructure programme nobody requested. Meanwhile the company’s knowledge stays exactly where it has always been, in people’s heads and in tools nobody has structured.
Where the data-team reflex comes from
The classic approach treats context as a data problem: extract, transform, index. That reflex was learned on analytics projects, where it is justified, and it gets applied out of habit to a problem that does not resemble them.
Because what an AI must know to be useful - who decides what, how the organisation works, which clients it serves, which commitments it has made - is an organisational problem. That knowledge is not stored anywhere in a usable form. It is obtained by interviewing the people who hold it, then having it signed off by the people who answer for it. No pipeline will ever extract who arbitrates commercial discounts, because that information lives in no database.
How your Kastel takes shape
Kastel starts from that observation and turns the method around. A short interview gathers from you how the organisation works. From your answers, Kastel proposes a structure, which the people in charge correct and then approve, domain by domain: nothing becomes authoritative without a human signing it off, and that is the very principle of context governance.
Your existing sources are not migrated; they are attached. Connectors link your storage, your email and your CRM to the context, right where those tools already live. The goal the product sets itself - and we state it as a goal - is that a context takes shape in one working session, not in a quarter. The product page walks through the method. As for what that context then changes in the answers you get, we covered it in why a connected AI answers wrong with confidence.
The share of the work that stays yours
This deserves saying plainly: Kastel does not remove all effort, it moves the effort to where it is useful. Your organisation must name an owner for each domain, approve what the proposed structure says, and respond when a fact comes into doubt. Kastel structures and governs; it does not invent authority. An organisation where nobody wants to be responsible for anything will not have a good context, with or without a data team.
That part of the work cannot be outsourced, and it is a blessing in disguise: it forces you to clarify who decides what, which pays off well beyond AI. If the question you are left with is whether to build the whole thing yourself rather than start from an existing base, we have put the two side by side in build your own context layer, or start from the free core. Before opening your tools to an AI, a few questions are worth asking first, and we have gathered them in what to check before connecting an AI to email, storage and CRM.
The test costs neither budget nor consultants
You can put this claim to the test without trusting us and without engaging anyone. The core of Kastel installs for free on your own infrastructure, through the command line and MCP, with no size limit. Run the interview, let the people in charge approve the structure, plug in your own AI keys, and judge for yourself whether a useful context has taken shape without a data engineer touching anything.
And if the answer is no, full export hands everything back in a readable format: you leave with your context, not with a debt.