The scan

The AI Accountability Scan

Know what your AI does, be able to prove it, and see where the work may go faster.

Four fields, one standard.

Education and learning materialsConsulting and strategyPublishers and editorial teamsResearch and think tanksTeachFoundryBookaflowValoraEducation and learning materialsConsulting and strategyPublishers and editorial teamsResearch and think tanksTeachFoundryBookaflowValora

The question

Sooner or later, someone asks.

A board, a procurement team, a funder, a client, or your own people: how do you use AI, and how do you know it holds up?

Most organizations have no answer today. Not because they're careless, but because nobody ever told them what there was to arrange. The scan exists so you can answer that question before it's asked.

And there's a second question, one almost nobody asks when buying AI even though it may matter most: in five years, will this system have made your people more capable, or more dependent? We cover that too.

What we assess

Three things decide whether your AI holds up.

The six layers of TRACES.

A trustworthy AI system isn't a clever model with a text box. It's the layers around it: where the answers come from, what the system may claim, what gets filtered, who may do what, what gets recorded, and who signs off before anything counts. We hold every AI system you use against those layers and rate each one: present, partial, or missing. Every AI incident you've heard of, from the airline chatbot that promised a discount that didn't exist to the newspaper with invented quotes, traces back to exactly one such missing layer. We use that lesson to point, not to frighten.

The evidence.

What you can show today and what's missing: an AI use policy, an overview of which systems you run and what for, supplier agreements about your data, the privacy assessment where one is required, and proof that your people know what they're doing (that obligation already applies). Measured against what your world actually gets asked for: education, research, publishing and consultancy each have their own requirements, and we know them.

The people.

Does sign-off actually happen, or is someone ticking boxes without looking? Does the system keep your people sharp, or is it quietly taking over the thinking? This is the part of the assessment almost nobody does, and for an organization that lives on thinking, perhaps the most important.

Want to try the yardstick yourself first? The free supplier scan walks you through these six layers for any AI system, no account needed.Take the free supplier scan

What you get

A report in plain language, no jargon.

Per systemTRACES

TTrailmissing

RRetrievalpresent

AApprovalpresent

CConstraintspartial

EExpertisepresent

SStoragepartial

which layers hold, which are missing, and the concrete risk behind every gap.

Per system

which layers hold, which are missing, and the concrete risk behind every gap.

The evidence checklist

what you have, what you lack, what it costs to put it right.

A prioritized repair plan

what comes first (biggest risk, smallest fix), what comes later, what you can leave alone.

The referral list

the points that need legal advice, each with the exact question to put to your lawyer.

The plan is yours, whether or not we carry on together afterwards.

How we build afterwards

Compliance isn't a plaster stuck on afterwards.

If the scan shows something needs building or repairing, it happens against the same standard we assess by. Everything we build is set up from day one to fully meet European and Dutch rules. Not patched on afterwards, but as a design principle:

Every answer shows its source. The system may claim nothing it cannot trace. If it can't support an answer, it says so.
The hard limits sit outside the AI. Calculating, access, money, publishing, deleting: never the model's job. Fixed rules do that, and rules can't wander.
A human signs off before anything counts. And the system is built so that sign-off stays real: it shows its uncertainty, so your people keep paying attention instead of ticking boxes.
Everything is recorded. Which source, which answer, who approved it. Not because you must, but because you can then show it when someone asks. That record is also a large part of your legal obligation, handled.
Your data stays yours. Clear agreements about where data lives and what will never happen to it. Within European rules, in writing.

That way compliance isn't a cost line next to the system. It's a property of it. The system you get is the proof.

The limit, said out loud

The scan assesses technology and organization.

It is not legal advice and not a certification. Where a legal judgment is needed, the report says so explicitly, including the question to put to your lawyer. After the scan you know exactly what's settled and what's still open. Drawing that line honestly is part of the product.

And there's work we turn down: systems that assess, select or admit people fall under the strictest European regime and belong with specialists. If that's where your question sits, we'll say so in the first conversation.

Price

Fixed price, known up front, no billable hours.

Compact

one AI system, the layer assessment and the evidence verdict, with a repair plan.

Full

every use of AI across the organization, including the people assessment and a presentation to your board or team.

The exact price depends on scope and you'll hear it in the first conversation, before anything starts.

Start small, go far.

Want to look for yourself first? The supplier scan and the Chain check are free and need no conversation.

The supplier scan

Want to know where you stand?

In half an hour we walk through what you use today and whether the scan is worth it. Sometimes the answer is: you already have this in order, or this is too small to assess. You'll simply hear that.

Book an intro call

Here's what that looks like in practice.

Three different worlds, one principle: the AI does the work, the human checks and approves.

Read the case