Most AI projects fail on the boring part.
Not the model. The data nobody cleaned, the workflow nobody mapped, the team nobody trained, and the handover nobody wrote. We do that part — and we hand you something you own and can actually run without us.
Four services. One of them is usually enough.
They overlap on purpose — most engagements start as one and quietly become another. We say which one you actually need before we quote, including when the answer is that you need none of them yet.
Read across the row you recognise.
The same eleven requests come up again and again. This is which service answers each one — and where the honest answer is that a service will only get you part of the way.
Four questions, then a scoped sheet you can print.
It runs entirely in your browser — nothing is sent, nothing is stored, and there is no email box at the end of it. What you get is the same first-pass scope we would write after a first call, so you can see the shape of the work before you talk to anyone.
What we would build
How it would run
What we would need from you
Shape of the engagement
The honest risk
The catalogue, with its real status showing.
These are our own tools rather than client builds. The first six run today on the Exzellent platform — Exzi is open to anyone, the rest sit in your dashboard once you sign in. Below them, what is being built and what is only planned, because a catalogue that hides the difference is a catalogue nobody can plan around.
Five steps, and you can stop after any of them.
Each step ends in something you keep — a document, a working prototype, a trained team. No step requires the next one, and nothing is designed so that leaving costs you the work already done.
The questions your legal team will ask, answered first.
In Germany this is not a footnote — it is usually the reason a project is approved or killed. So it is on the page rather than in a follow-up email, and it is written to be forwarded to whoever has to sign.
What we won’t build, whatever the budget.
Some of these are law, some are craft, and one is just a judgement we have made and will defend. Better to know now than after a kick-off meeting.
Fixed before we start, or we haven’t scoped it well enough.
Open-ended day rates on an AI project are a way of moving discovery risk onto the client. We would rather do the discovery properly, quote a fixed number, and be wrong at our own expense.
The ones that decide the project.
Tell us the problem, not the technology.
The best first messages describe something that is slow, expensive or error-prone today. We will tell you whether AI is the right instrument for it — including when it is not, which happens often enough to be worth saying out loud.