Versatil
For the person who signs

Find out what your operation actually decides.

Four weeks, one team, one real decision cycle. We arrive in your building, read the work you already hold, and run your next committee inside the system. You keep the record it produces, and the twin of your own house.

Book the auditSee what it produces
Fixed price, fixed scope, fixed end date. If you stop there, you still keep everything below.
The cost of not knowing

Ungoverned, it already costs you. There is just no budget line with that name.

Nobody provisions for this, which is exactly why it grows. Every figure below is someone else's published measurement, attributed, not ours.

2 068 $
spent per employee per year on AI in 2026
30 to 50%
of that budget lost to shadow duplication, wrong-sized models and unused licences
4.63 M$
average cost of a breach involving shadow AI, of which 670 k$ is the shadow AI itself
What that arithmetic gives, for a thousand people
AI budget2.07 M$
Waste already inside it620 k$ to 1.03 M$
Likely overrun, since 63% of companies exceed by 30% or more in year one~620 k$
Leaking every year, with no line carrying the name1.2 to 1.7 M$
Then it changes nature

Today it costs you a leak. Next it costs you the programme.

17% of organisations have deployed agents today. 64% will be in production by 2028, the fastest adoption curve Gartner has ever recorded. And their verdict on where that lands: more than 40% of agentic projects cancelled by the end of 2027, for escalating costs, unclear business value and inadequate risk controls. They name the mechanism: agent sprawl.

01Scoped, graduated autonomy
02Human verification gates, calibrated to the stakes
03Per phase ROI checkpoints, signed off by finance
04A named governance owner, per agent

That is Gartner's list of what the surviving projects do, not ours. It is also, point for point, what the audit installs. An analyst your board already pays wrote the specification.

Figures from published 2026 industry research, cited as theirs. Some come from vendors who sell AI spend management, so treat them as orders of magnitude, not as our measurements.
Why this shape

You already buy diagnostics. You do not yet buy AI platforms.

Most enterprise AI stalls in procurement, because a platform is a new category and a new category takes two quarters and a committee. An audit is a line item that already exists in your budget, with a beginning and an end. The difference is what it leaves behind: a consulting engagement leaves a document, this one leaves the operating memory of your own house.

4
weeks, one team, one real decision cycle
1
purchase order, on the diagnostic line you already have
0
permanent systems to security review before you start
The four weeks

It arrives, learns your operation, runs your cycle, and answers for it.

Nothing here is a workshop. The only thing that happens is your real work, done in the system, with the reasoning kept as it goes.

Before
The five questions, answered on your own installation

Sovereign, private cloud or self-hosted, your keys, scoped ingest. You also choose the model, because it is yours to choose.

Week 1
Scan my work

Each participant points it at their own work. It reads what is already there and proposes projects, goals and teams it found. You approve or you skip. Nothing is applied on its own.

Week 2
Your profession, configured

The twin ranks the Domain Packs against your workspace and proposes one. You confirm it. The Atelier is seeded for the way your field actually works.

Week 3
Your real cycle

Your committee meets inside the product. Calls arrive with a conviction score and the objection attached. You decide exactly as you always did. The difference is that it deposits.

Week 4
Replay, and decide

We replay one decision in front of you: the choice, the why, the sources, the dissent that was kept. Then you keep it or you delete it, and the deletion is as inspectable as the record.

What you keep

Six things, and you keep them whether you continue or not.

01The twin of your house, seeded with your operation, running in your environment.
02The decisions from a real cycle, on record and replayable.
03The Domain Pack configured for your profession.
04The map of how work is actually distributed: who decides what, where it queues, where it stalls. This one has value on its own.
05Your cross-person reuse baseline: how much of one person's work already rests on someone else's deposit.
06The five trust answers, demonstrated on your own installation rather than promised in a deck.
Trust, answered concretely

Five questions, on screen, before you sign.

Who owns the data?
You do. It stays in your environment.
Where is it cached?
Inside your perimeter, shown to you.
Are the prompts protected?
Scoped per tenant, demonstrated live.
Does anything reach a vendor?
No vendor-side copy. Sovereign by default.
Can you replay the why?
That is the whole point of the record.
Price

Stated plainly, credited in full.

One line, fixed. No per seat, no per call, no per token during the audit, because charging for usage would tax exactly the accumulation this is meant to produce.

45 000 €
fixed · 4 weeks · credited in full against year one
One team, one decision cycle, deployed in your environment or on our hosted surface, your choice of model.
75 000 €
sovereign · air-gapped or on your own infrastructure
Includes the security review and the five trust answers demonstrated on your infrastructure. A sovereign deployment genuinely costs more to run, so it is priced as what it is.
Versatil is pre-revenue and says so. Every screen you will see before your own data lands is specimen data you can open.

Four weeks from now, you can show your process instead of describing it.

One team, one cycle, one purchase order. You keep what it produces either way.