An independent three-week review: what you run, what's worth keeping, and which EU rules already apply to it.
Takes about a minute. No email required. Or book a 30-minute call.
Pick the system closest to yours. Written twice — once for the board, once for engineering.
An orientation, not a legal determination. What actually applies is what the review settles.
High-risk duties moved to 2027 and much of the market read that as a pause. The obligations reaching ordinary mid-sized companies didn't move at all.
We work out which apply to your systems. Where a formal legal opinion is needed, we introduce qualified counsel by name.
Each stage leaves you something you own and can act on without us. Advice you can't walk away from isn't advice, it's a dependency.
Three weeks. We map every AI system in the organisation, judge which are worth scaling and which should stop, and hand over a prioritised plan.
We finish the cases worth building, with the oversight around them. On the systems you already have.
Systems drift, vendors revise terms, teams add tools. Monthly review of what runs against what was agreed.
Interviews across departments, tool and contract review, mapping where data travels.
What applies to each system, where oversight is missing, what your vendors are doing.
Written report, prioritised remediation, a plan naming who does what by when.
Ninety minutes with leadership, and a separate session for the people using the tools.
For the first two companies: €2,500 instead of €9,500, focused on two departments over ten working days. In return, honest feedback and a written reference if the work was worth it.
Nothing about the findings changes. A report that is easier to sell is a report nobody should buy.
Ask about the founding rateHalf the review fee is credited against implementation starting within 60 days. The other half stays with the review, which is priced to stand alone — a review that pays for itself only when you build is a review you should distrust.
Not a document for a drawer. The practical difference, in things you can check yourself.
Every plan splits three ways. If we billed for all of it, the review would stop being honest.
Most AI advice comes from someone holding a product — a partnership, a margin, a licence to move. The recommendation and the revenue point the same way.
It's why we publish fixed prices, and why every plan says plainly which parts are ours.
No, and we say so before anyone asks. We establish which obligations apply to which systems and what to do about them. For a formal legal opinion we introduce a qualified lawyer by name.
Nobody can, and be wary of anyone who says otherwise. What you can have is a reasoned position on every system and evidence that you acted on it — which is what auditors and enterprise customers actually ask for.
They work on the legal position. We work on the systems, so that position is true in practice. Most companies have someone who can read the regulation and nobody who can map it onto the tools staff actually use. That gap is the job.
The duties already in force have no size exemption, and smaller companies tend to have more unmanaged use, not less — nobody's job is to notice. If your exposure is smaller than you feared, the report says that too.
Then you found out from us rather than from a candidate, a customer or a regulator. Findings are sequenced by what's urgent and what's cheap, and nothing in the report leaves your organisation.
Supvisor is a new practice and won't pretend otherwise — hence the founding rate and the absence of logos. What isn't new is the work: building AI systems that run in production, and knowing where they go wrong when nobody is watching. If you want a long client list, we're the wrong firm this year.
We build. The review decides what's worth building; most of what follows is taking those cases to production, with approval steps and records built in. We are paid for the work, not for how much AI you end up running — which is why the review is as willing to say stop as build.
Five to eight interviews of about 45 minutes, your tool and contract records, and one workshop per department. Beyond that, the three weeks are ours.
Thirty minutes, no preparation needed. Tell us what you're running and we'll tell you plainly whether there's a problem worth paying to solve.
Or email contact@supvisor.ai
We'd rather say that plainly than fill the page with badges we haven't earned. The first two engagements run at the founding rate.
Most decisions here need more than one person.
