Your staff already use AI. The question is whether you can explain how.
Article 4 of the EU AI Act requires every organisation that uses AI to ensure demonstrable AI literacy among its people. Including organisations that only use ChatGPT or Microsoft Copilot. I train your team on their own work, with their own documents, and deliver the file that shows you have it covered.
Schedule a trainingEnforcement began on 2 August 2026
The obligation itself has applied since 2 February 2025, but until recently nobody could act on it. That has changed: since 2 August 2026 the Dutch authorities can actually enforce it. In the Netherlands coordinating supervision sits with the Dutch Data Protection Authority and the RDI, through the Algorithm and AI Chamber. One misconception worth clearing up: Article 4 does not target companies that build AI, but everyone who uses it. If you have rolled out Copilot, or your people use ChatGPT for quotes and emails, you fall under it. The law prescribes no exam and no certificate — but it does require the level to match the role, and that you can show what you have done.
No certificate required. Evidence is.
Untrained staff cost more than training does
The reality across smaller organisations is almost always the same: there is no policy, so everyone improvises. That creates two problems that each cost money on their own.
Shadow IT
Without clear boundaries, people paste client data, quotes and personnel files into whatever free chatbot they found. Not out of malice — because nobody taught them anything better and the work has to get done. That is a data breach in the making, and you hear about it only once it goes wrong.
Expensive licences sitting idle
The other side: Copilot licences nobody really uses because the introduction consisted of an email with a manual attached. Forrester measured a 353% ROI and an average of nine hours saved per user per month in the SMB segment — but only at organisations that introduced it deliberately.
Output nobody checks
A model that sounds convincing and is occasionally wrong is more dangerous than one that clearly does not work. Anyone who does not understand how a language model reaches an answer will not check it either. That is precisely why the law demands literacy rather than a user manual.
From inventory to demonstrably literate
AI readiness session
A half-day working session with the people who do the work. We map where AI is already being used — including the tools you did not know were in the building — and walk through the processes where it holds promise or, conversely, risk. This doubles as the inventory Article 4 asks of you.
Boundaries and policy
Per process step we record which data a model may see and which must never leave the organisation. That becomes an internal policy in plain language that people will actually read: one page covering what is allowed, what is not, and who to ask when in doubt.
Training per role
A director deciding on AI investments needs something different from an administrator processing invoices or someone answering the phone. I train in small groups on the work they actually do, using documents from their own practice. Not a general AI presentation.
Record-keeping
You take the file with you: the inventory, the policy, and the overview of who completed which instruction and at what level. That is what you produce when someone asks how you meet Article 4 — not a certificate, but a demonstrable approach.
Who is this for?
SMEs with 5 to 250 employees
Large enough that people start using tools on their own initiative, too small for a compliance department to pick this up. Exactly where the gap is widest: 13.8% of Dutch micro-enterprises used AI in 2025 against 66.2% of large enterprises.
Practices bound by confidentiality
Accountancy and advisory firms, healthcare, legal. Where the question of what may and may not enter a model is not a theoretical debate but a daily judgement with professional consequences.
Organisations that just bought Copilot
The licences are rolled out, usage lags behind and nobody knows whether it is happening safely. Focused training recovers the investment and covers the statutory obligation at the same time.
Foundations and non-profits
Tight budgets, heavy accountability requirements, often volunteers handling personal data. Half a day of clarity goes a long way there.
Implementing Agentic AI under the EU AI Act.
The practical implementation guide for organisations deploying AI agents on processes that genuinely matter. No legal treatise — what you must arrange, in what order, and why that order matters.
- A tick-box implementation checklist per AI application
- Ready-made templates for the AI register and validation log
- A system instruction against prompt injection you can adopt directly
- The full procedure for conformity assessment and CE marking
Get the guide
Enter your email address. You will receive a confirmation mail, and the guide is yours straight after.
European hosting, European rules.
What you learn in the training matches how we build: risk classification up front, human oversight built in, logging from day one. Only someone who understands the system and its limits can supervise it — which is why training is not a side product here but the first step.
Can you demonstrate that your staff are AI literate?
If not, that is sorted within half a day. Schedule an intro call and I will tell you what your situation needs — and what it does not.
Schedule a training