AI training plan (Article 4 of the AI Act): the how-to

The essentials in brief

Article 4 of the AI Act requires any organisation that uses AI to ensure a sufficient level of AI proficiency ("AI literacy") among the people who deploy or supervise it. This obligation has applied since February 2025, but the supervision and penalty regime that comes with it becomes fully operational on 2 August 2026 — which suddenly makes it very concrete for businesses.

The good news: nothing mandates a single format or an official certification. The law calls for a level of competence suited to your uses, roles and risks. It is up to you to build it. This article sets out the method to build a training plan that meets the obligation and genuinely serves your teams.

What Article 4 says (and what it does not)

Article 4 requires providers and deployers to take measures to ensure a sufficient level of AI proficiency among their staff, taking into account their knowledge, their context of use, and the people affected by the systems.

What the text does not require: a diploma, training from an accredited body, or a set number of hours. What it does require: that the people who use AI understand what they are handling, its limits, its risks, and know when to alert a human. Compliance is demonstrated by a structured, documented plan — not by a certificate.

Why "one generic course for everyone" is not enough

The classic mistake is to order a single, one-size-fits-all awareness session and tick the box. It misses twice:

• It over-trains those whose use is marginal and under-trains those who make AI-assisted decisions.
• It does not connect the training to the real uses of the business, so it does not reduce concrete risks.

Useful AI literacy is differentiated: the level expected of an executive, of an HR manager using AI to screen applications, and of an employee drafting with an assistant is not the same.

Building your plan in five steps

1. Map AI uses

Where is AI actually used in the business, by whom, and for what? This mapping is the common foundation of AI Act compliance — it also serves to classify uses by risk level. We detail the method in framing AI within your business processes.

2. Define proficiency levels by role

To each group, assign an expected level: basic awareness for everyone, stronger skills for those who deploy or supervise sensitive systems, a governance culture for decision-makers.

3. Build tailored paths

A common base (understanding AI, its limits, the risks, the internal rules) complemented by targeted modules by role. Short, concrete, anchored in your real tools rather than theoretical.

4. Bring non-technical staff on board

This is often the neglected link. The people least at ease with AI are the ones who get the least from it — and who make the most oversight errors. A good plan puts them at the centre, in jargon-free language.

5. Document it and keep it alive

Keep a record: who was trained, on what, when. That is what demonstrates compliance. And plan for regular updates: uses and tools evolve quickly.

Literacy, the foundation of any AI project

Beyond the obligation, building AI awareness is the prerequisite for any AI project that works. Even an excellent tool fails when teams understand neither what it does nor where its limits lie. Investing in AI proficiency is not only about compliance — it gives your next AI projects the conditions to succeed.

That is why we treat training not as a box to tick, but as the first step of a broader AI-applied-to-your-processes approach. The mapping that grounds the training plan is the same one that underpins your AI Act compliance as a whole.