AI system & product overview
What the system is, what it does, which AI capabilities it uses, and where it's deployed.
EU AI Act · Transparency & disclosure obligations
A guided assessment that turns what you already know about your AI product into a professional Transparency Evidence Pack — the document your enterprise buyers, security reviewers and auditors keep asking for.
“A prospect sent us a 40-question AI due-diligence form. We had nothing to send back.”Founder, Series A SaaS
“Our AI disclosure lives in three different documents and none of them agree with each other.”Head of Product
“We're probably fine. The problem is we can't prove it in writing to anyone who asks.”Compliance Lead
Most teams don't have a transparency problem. They have a documentation problem. You already disclose, oversee and limit your AI — it's just never been written down in one place that a buyer or regulator would recognise.
What you get
Your answers are turned into clear, professional statements — not a dump of form fields. Every claim is tagged with where the evidence lives, and anything missing becomes a tracked action.
What the system is, what it does, which AI capabilities it uses, and where it's deployed.
Who it's for, who is affected by its outputs, and — just as important — what it is explicitly not for.
Plain-language statements of how and where people are told they're interacting with AI, and how generated content is marked.
Your oversight model, who reviews what, how outputs get overridden, and the route for someone to contest a result.
Documented failure modes, where performance drops off, the controls you run, and the risks you knowingly accept.
What goes in, whether it's personal data, whether it trains anything, where it's processed, and how long it's kept.
The actual notice wording, where it appears, in which languages — plus drafted templates you can lift straight into your product.
A control register: what's in place, who owns it, how often it's reviewed, and what triggers a re-review.
A line-by-line table of what information was provided and where it can be verified. This is the page reviewers read first.
Everything you couldn't answer, ranked by severity, each with a concrete next step and an owner field.
Up to six notices written from your own answers — in-product labels, a first-contact disclosure, an AI-generated content line, your primary AI notice, a privacy-notice section and a responsible-use note for business customers. Copy and publish.
Oversight, disclosure, data handling and limitations on a single page. Send it to the prospect who asked one question, and keep the full pack for the people who want all of it.
Duplicate last release's pack and the new one records exactly what changed, answer by answer — so re-issuing is a review, not a rewrite.
How it works
Nine short sections, in plain English. Every question tells you why it's being asked and what a good answer looks like. Don't know something? Skip it — it becomes a tracked gap, not a blocker.
~45 minSee an evidence readiness score, plus every gap ranked Critical, Important or Recommended — each with a specific action, not vague advice. Fix what you can in the moment, assign the rest.
~10 minYour answers become a formatted document with a cover page, statements, control register, evidence checklist and gap register. You also get your user-facing notices drafted from your own answers, ready to paste into the product.
InstantSave as PDF, download as Markdown or self-contained HTML, or export your structured answers so you can update the pack next release instead of starting over.
~2 minWho this is for
“This is blocking a deal and I have two hours.”
You need something credible in front of a buyer this week. You'll answer most of it from memory, flag the rest, and export.
“I know the product. I don't know what they want written down.”
You have all the answers already. The value here is the structure, the wording, and knowing when you're done.
“I need a defensible record, not a marketing page.”
You get a control register, an evidence checklist with source references, versioning fields and a gap log you can actually work through.
Straight answer
We help you prepare the evidence. Your legal counsel decides what it means.
Pricing
Start the assessment free. You only pay when you're ready to export a finished pack.
$99one-off
One product. Everything you need to answer a buyer this week.
$249one-off
For teams shipping more than one AI feature.
$499one-off
For a portfolio, an agency, or a compliance function.
Every plan includes the full questionnaire, the gap review and the pack on screen — those are free and always will be. You pay to export. Your own answers export as JSON free, whatever you choose. Checkout runs on Stripe; prices exclude VAT.
Questions
No. This is an evidence preparation tool. It helps you assemble and articulate what your product already does in a format that reviewers recognise. It does not interpret the law for your situation, determine your risk classification, or tell you whether you are compliant. If you need that, take your finished pack to qualified counsel — you'll get a much faster, cheaper answer than starting from nothing.
Most people finish in 30–60 minutes. If you know your product, the questionnaire moves quickly — roughly half the questions are yes/no or multiple choice. The longest sections are limitations and human oversight, because those are the ones worth thinking about. You can leave and come back; progress saves automatically.
Skip it. Unanswered questions become tracked gaps in your pack, ranked by severity, each with a recommended action. A pack that honestly says “we haven't defined this yet, here's the plan” is far more credible than one with confident answers nobody can back up.
Yes, and it's one of the most common situations. If you build a product on someone else's model, you're still the one talking to your users — so disclosure, oversight, limitations and notices are still yours to document. The questionnaire branches based on whether you build the model, use someone else's, or both.
No. Mark anything that genuinely doesn't apply as not applicable and say why. Declared questions are excluded from your readiness score rather than counted as failures — and they're listed in the pack with your reason, so a reviewer can see the decision. They're also excluded from the score rather than counted as satisfied, so waving things away can't inflate the number: if you declare a large share of the questionnaire, the headline says so.
In this version, nowhere. Your answers are stored in your own browser and the pack is generated locally on your machine. You can export a structured copy at any time, and clearing your browser data clears your answers — so export before you do.
Generally yes — one pack per AI system, because intended purpose, oversight and limitations differ between them. Your workspace holds as many packs as you need, side by side, each with its own readiness score and gap list. If two features genuinely share a purpose, an oversight model and a set of limitations, one pack covering both is fine — just say so in the scope statement.
That's what the version field and pack duplication are for. Copy last release's pack, clear the version, update what changed, and export — the old pack stays intact as a record of what was true then. Most teams re-run it at a major release, when they change model provider, or when they enter a new market.
The pack includes an accountable owner field and space for a reviewer. In practice that's usually the product lead who knows the system, countersigned by whoever owns compliance or legal. The point isn't the signature — it's that one named person can answer questions about it.
Start the assessment now — you'll know within five minutes whether this is worth your afternoon.