Your next enterprise review will ask how your AI works. Have the evidence ready.
We inventory the systems in your product and operations, map provider and deployer roles, identify evidence gaps, and prepare reusable answers for Procurement and Legal. Fixed scope, fixed deadline, delivered in English.
Art. 4 · AI literacy
We cover thisIn force · since 2 February 2025
If your staff use AI, you must take measures to support their AI literacy, proportionate to the risk and the role. The text no longer asks you to guarantee anyone’s level: it asks you to act, and to be able to show that you did.
The Omnibus rewrote it: it was "ensure a sufficient level" and is now "take measures to support the development". Still binding, but met through documented measures rather than a per-person level.
- Art. 5Prohibited practices
In force · since 2 February 2025
Eight uses of AI are banned outright, including inferring a person’s emotions in the workplace.
- Art. 4AI literacy
In force · since 2 February 2025
If your staff use AI, you must take measures to support their AI literacy, proportionate to the risk and the role. The text no longer asks you to guarantee anyone’s level: it asks you to act, and to be able to show that you did.
The Omnibus rewrote it: it was "ensure a sufficient level" and is now "take measures to support the development". Still binding, but met through documented measures rather than a per-person level.
- Chapter VGeneral-purpose AI models
In force · since 2 August 2025
Binds whoever places a model on the market, not whoever uses it. Relevant if you train or fine-tune your own.
- Art. 50Transparency
In force · since 2 August 2026
Anyone talking to your chatbot must know it is an AI, and synthetic content you publish must be labelled as such.
- Annex IIIHigh-risk systems
Deferred · from 2 December 2027 · was 2 August 2026
Recruitment, credit, education, biometrics. Point 4 covers employment — the one that catches HR software.
- Annex IAI in regulated products
Deferred · from 2 August 2028 · was 2 August 2027
AI embedded in products already under EU product-safety law: machinery, medical devices, toys.
Dates per Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744.
What changed on 24 July
The Digital Omnibus was published in the Official Journal on 24 July 2026 and entered into force on the 27th, six days before the date it was deferring. Almost everyone stopped at the deferral headline and missed the second thing it did.
Deferred
- Annex III — standalone high-risk systems, recruitment included.
- Annex I — AI embedded in products already under EU product-safety law.
Rewritten
- Article 4 — AI literacy moves from "ensure a sufficient level" to "take measures to support the development". From an obligation of result to an obligation of means.
- The new text clarifies that it does not require guaranteeing any given individual’s level. It still binds every provider and deployer.
Unchanged
- Article 5 — prohibited practices, no exceptions, no grace period.
- Article 50 — transparency and synthetic-content labelling.
- Chapter V — obligations on general-purpose model providers.
The deferral changed the regulatory calendar, not your customer’s due-diligence timetable. Procurement teams can still ask for a systems inventory, roles, controls and written answers before they approve a supplier.
Read the regulation on EUR-LexRegulation (EU) 2026/1744 · OJEU 24 July 2026
Give Procurement and Legal something they can review
We do the technical work behind the answers: system inventory, preliminary classification, evidence gaps, Article 4 measures and Article 50 notices. Each engagement ends with a defined set of documents.
- Annex III→ 2 December 2027AI systems auditWhich AI your company runs, under which legal role, and what risk class each system falls into.What you get: Systems inventory + risk classification + gap report7–10 working days
- Art. 42 February 2025AI risk trainingRole-based tracks and an internal record of the measure taken, without presenting training as a compliance certificate.What you get: Role-based programme + tracking platform + per-employee record2 weeks to rollout
- Art. 502 August 2026Transparency complianceAI-interaction notices and synthetic-content labelling, in force since August 2026.What you get: Touchpoint inventory + notice copy + labelling guide5 working days
- Annex III→ 2 December 2027Documentation & RFP answersThe answer matrix your sales team needs when procurement asks about the AI Act.What you get: Answer matrix + system fact sheet + supporting file7–10 working days
Keep a record of the measures you took
Article 4 does not require a certificate. You do need to explain which measures you chose, for which roles and when. The platform records each person’s track, completion date and check result, then produces a leadership view.
- Workforce dashboard
- Coverage by department and by track. Who is missing, since when, and who needs a nudge.
- Per-employee record
- Assigned track, completion date and check result. An internal record of the measure taken.
- Completion record
- Per person, dated, listing what was covered. Exports to PDF and CSV for the internal file.
AI literacy · Art. 4
Workforce coverage 82% · 80/98
| Department | Completed | Status |
|---|---|---|
| Sales | 24/24 | Complete |
| People | 9/9 | Complete |
| Operations | 31/41 | In progress |
| Engineering | 12/18 | In progress |
| Leadership | 4/6 | In progress |
Demonstration with fictional data. The platform ships as part of the training service.
A scope your team can approve
The scope, price, delivery date and document set are agreed before work starts. No open-ended consulting engagement.
Scoping call
20 minutes to see what AI you run and what you are being asked. If you need nothing, we say so there.
Written scope and price
In writing, with what is in and what is out. No mid-project scope surprises.
Delivery and walkthrough
You receive the document set and a final walkthrough. Any unresolved point stays explicitly marked as unresolved.
Questions
Typically: a systems inventory, your provider or deployer role, preliminary risk classification, the controls already in place, evidence gaps and reusable answers for their questionnaire. We turn those questions into a reviewable document set.
Is a customer waiting for answers?
In twenty minutes we can review the questionnaire or due-diligence request and tell you whether a fixed-scope engagement fits.
If you need none of these services, we will tell you on the call.