Skip to content
I'm here as…

For compliance and operations leads

See what may apply. Show the work behind your answer.

EU AI Fit takes each AI system through five steps: Describe, Assess, Plan, Operate, Prove. Each duty that may apply gets its own item, owner and evidence, so you can show the work behind every answer.

  • Asks first whether the Act applies to you
  • Lists each duty as its own item, with the evidence expected
  • Keeps every approval and reason on record

Try it: a three-question exposure check

Runs in your browser
What does the AI system do?
How is it provided?
Do users know they are dealing with AI?

Early screen · 1 flag

  • Third-party dependency needs assurance
  • No obvious trigger found in this early screen

Suggested next stepAsk the supplier for the information you need, check whether you are the provider of your own system, and set a reassessment date.

Run the full free check

The answers start on an example, so change them to match your system. This shorter version uses the same rule as the full five-question check, which asks more. It is an early screen, not a legal classification or legal advice.

Next action · Example data

Assign evidence for human oversight

Customer Reply Assistant · Plan

Morgan Reed · Owner

What you'd use

The Workspace

Home, AI systems, evidence and reports in one place, with the next action always in view.

  • Workspace
  • Learner
  • Adviser
  • External reviewer

Five steps · Example data

3of 5 · Plan

Customer Reply Assistant: described, assessed, now being planned.

  1. Describe
  2. Assess
  3. Plan
  4. Operate
  5. Prove
See all five steps

Ways to start

from£0

The free exposure check screens one AI use case. No account required.

See the three ways to start

What to expect

  • Source-linked guidance
  • Independent review on request
  • No certification claims

Software that records screening results, duties and evidence. Not certification, regulatory approval or legal advice.

Why now: some duties already apply.

Dates from the current text of the Act · Full timeline

Articles 4 and 5

Since 2 February 2025

Banned practices, and AI literacy measures for staff

Article 50

Since 2 August 2026

Transparency duties, such as telling people they are talking to AI

Article 5

From 2 December 2026

Two more uses banned, including non-consensual intimate deepfakes

Annex III

From 2 December 2027

High-risk rules for the uses listed in Annex III, such as recruitment

Annex I

From 2 August 2028

High-risk rules for AI in regulated products

Room 2 · The path

From an AI question to an evidenced decision.

The service leads each system through the same controlled path. It shows what is known, what still needs judgement and who must act next.

Step 1 · Describe

Record what the AI system does and name an accountable owner.

What you do
Record what it does and uses; nominate an owner.
Record it leaves
A system record with an accepted owner.
Who approves
The owner accepts.

Exposure map, scope and role, and classification, each approved by a person.

Select to see the record it leaves

Each duty becomes its own item, with the evidence it expects.

Select to see the record it leaves

Gated lifecycle moves, risks, incidents, suppliers and staff briefings.

Select to see the record it leaves

Readiness report, disclosure pack, documents and independent review.

Select to see the record it leaves

Room 3 · See it working

Product demonstration · 2 minutes 56 seconds

What EU AI Fit does

See how EU AI Fit connects the AI inventory, accountable owners, assessment, duties and evidence, operation, training and reporting in one maintained record. All organisations, people and systems shown were created for demonstration, and no client data appears.

Watch on its own page with the full transcript

Read the video transcript

EU AI Fit gives an organisation one practical place to govern the AI it builds, buys and uses. It is not a one-off assessment. It is a working record that stays current as systems change. Home starts with what needs attention now. The next action says why it comes first. Here, a stage change is due in two days and needs a decision. Home also shows where each system stands in the five steps of its record. My work gathers every open item across the workspace into one list. Each item gives the reason for its priority in plain words. There is no hidden score.

The AI systems list holds everything the organisation uses, whether it was added by hand, imported from a spreadsheet or created from an exposure check. Each system shows its stage, its accountable owner and where its review stands. Every system has one record, read in five steps: Describe, Assess, Plan, Operate and Prove. The record shows which step the system has reached, and what is still to do in each. In Assess, the team reviews the organisation's role and the system's classification under the EU AI Act. These are screening results that a named person reviews and approves. They are not legal conclusions.

In Plan, the approved role and classification become a list of duties. Each duty names the article it comes from, an owner, a target date and the evidence a reviewer would expect to see. Evidence stays attached to the duty it supports, stored privately in EU AI Fit or kept in the organisation's own systems. In Operate, moving a system into a pilot or into production is a decision, not a click. EU AI Fit checks the record first, and an owner or administrator approves or rejects the request.

Risks stay connected to the system, with their controls, the indicators the team watches, and any incidents. People are part of the record too. Staff are assigned AI literacy learning that fits their role, and what they complete is kept as a record of the measures the organisation has taken. In Prove, the record becomes a versioned buyer readiness report, which can also be sent as a frozen copy to an independent specialist for review.

Policy drafts are generated from the inventory, with the gaps marked for someone to fill in before anyone approves them. EU AI Fit does not certify compliance, and it is not legal advice. It gives the people who decide a clear record of what the organisation uses, who is responsible, what has been reviewed and what should happen next.

Room 4 · One product, four portals

Everyone gets their own way in.

Four portals onto the same record. Pick one and try it: the buttons inside work.

Example data

Who uses this: the compliance lead and colleagues, as Owner, Administrator, Contributor or Viewer.

Northstar Support Systems

⌘K Search
  • Home
  • AI systems
  • Suppliers
  • Risks & incidents
  • Evidence & documents
  • Reports
  • Training
  • Settings

Next action

Assign evidence for human oversight

Customer Reply Assistant · Plan · owner Morgan Reed

4

AI systems

7

Open obligations

3

Evidence items due

Work queue

  • Customer Reply AssistantPilot · transparency duties may applyPlan · evidence to assign
  • Ticket Priority ClassifierDiscoveryNeeds review
  • Help Centre SearchPre-productionMove to Production · due 16 Nov 2026
  • Recruitment assistantNot yet describedNeeds an owner

Room 5 · What’s on the desk

Six parts, one record

A working readiness process, not a one-off report.

AI inventory, exposure, obligations, evidence, staff learning and independent review sit in one controlled workspace, so each answer stays attached to the record behind it.

Every AI system follows five steps

  1. Describe
  2. Assess
  3. Plan
  4. Operate
  5. Prove
  • Inventory
  • Classification
  • Evidence
  • Suppliers
  • Literacy
  • Review

AI inventory

Capture what each system does, who supplies it, who uses it and which decisions or outputs it affects.

Exposure and classification

Scope and role questions, an exposure map, then five classification routes screened. A person approves each result.

Obligations and evidence

Each duty is its own item, with its article, owner and expected evidence, plus an Annex IV draft for providers.

Supplier assurance

A supplier questionnaire by secure link, evidence review, conditions, actions and a reassessment date.

AI literacy

Measures to support AI literacy (Article 4): role-based learning, system briefings and completion records.

Independent review

Send a frozen copy of the record by secure link. The signed review joins the audit trail.

AI literacy

Train the people who use, manage and govern AI.

A pathway for each role, briefings on the AI systems people actually use, a short check of understanding and a completion record, in the same workspace.

Explore AI literacy

Assign role pathways, brief staff on the real systems they use, check understanding and keep the record. The Act sets no level and needs no certificate.

  1. Assign a role pathway
  2. Brief on a real system
  3. Check understanding
  4. Keep the completion record

Free exposure check

£0

For an initial screen of one AI use case.

  • No account required
  • Potential risk and transparency flags
  • Suggested next actions
  • No definitive legal classification
Run exposure check

Readiness sprint

from£1,500

A guided pilot for B2B SaaS teams.

  • One-off pilot engagement
  • Inventory of up to 5 AI systems
  • Expert-supported classification workshop
  • Obligation and evidence plan
  • Buyer-ready readiness report (not a certificate)
  • Thirty-day pilot access
  • Then £450 a month to keep the workspace
Explore the sprint

AI literacy

£25/learner a year

Role-based AI literacy training on its own, for every member of staff.

  • Minimum 10 learners, billed yearly
  • Foundation and role-based courses
  • A completion record for every learner
Try the free taster course

For DPOs, consultants and advisers

Adviser pilot

Custom

Run every client from one place, with each client’s records kept apart.

  • A separate workspace for each client
  • Reports co-branded with your firm, approved by the client
  • One method across your whole portfolio
  • Limits and access set by each client's contract
Your client workspacesExample data
  • Northmoor Food Distribution4 AI systems · own workspaceReport ready to share
  • Solent Community Energy2 AI systems · own workspaceSupplier review due
  • Harbourlight Housing3 AI systems · own workspaceTraining 8 of 10 done

Each report carries your firm’s name once the client approves it.

Questions, and a quiet newsletter

More questions

Can I try before paying?

Yes. The public exposure check is free and requires no account. The readiness sprint is a paid, guided pilot: we set up the workspace with you and support the first classification decisions, so your team starts from reviewed results.

What counts as an AI system?

The Act's definition (Article 3(1)): a machine-based system that operates with some autonomy and infers from its inputs how to generate outputs, such as predictions, content, recommendations or decisions, that can influence physical or virtual environments. Record what you build, buy or use, then check each one against that definition.

Do you store my data?

The public exposure check works out its result in your browser. When you finish, the answers are sent to our server to be checked, and we keep only the result category and time, not the answers. If you ask for an emailed copy, your answers go into that one email and are not stored. Pilot workspaces can use reference-only evidence or an agreed managed private-storage tier.

Is this legal advice?

No. EU AI Fit is decision-support software. We help you organise, assess and document your readiness; final legal interpretation should come from a qualified adviser.

Newsletter

Practical updates, when there is something worth sharing.

Occasional notes on pilot places, product improvements and EU AI Act readiness. No weekly noise.