What is AI literacy?

AI literacy is the combination of knowledge, skills and understanding that lets someone use AI systems deliberately and recognise their risks, limits and consequences. It is not about programming; it is about knowing what a model can and cannot do, when the output needs to be checked, which data must never go into it, and when a human must make the decision.

The legal obligation

Why it matters

From understanding to demonstrability.

Until recently, AI literacy was a good idea. Since Article 4 of the EU AI Act, it is an obligation: every organisation that uses AI must ensure an adequate level among everyone who works with it on the organisation's behalf.

The practical difference lies in the word demonstrable. A successful afternoon with an engaging speaker changes nothing about your position when asked to prove it. A register with role mapping, test results and documented policy does.

What the EU AI Act actually requires →

The levels

Who needs which level?

Training everyone the same way is expensive and insufficient. The law expects the level to fit the role, the knowledge and the context.

Level 1

Basic

Everyone who uses AI tools

Recognising what a language model is and what it is not. Knowing what a hallucination is and how to spot one. Knowing which data should never go into a prompt. Understanding that the user remains responsible for the output.

Can use AI safely for their own work.

Level 2

Advanced

Anyone using AI in decision-making or customer contact

Assessing whether an output is usable, and on what basis. Recognising bias and cases where the model operates beyond its scope. Knowing when a result needs to be checked by a second person and how to record that.

Can weigh AI output and account for it.

Level 3

Accountable

Leadership, procurement, legal affairs, privacy officers

Assessing which AI systems enter the organisation and under what conditions. Applying the AI Act's risk categories to the organisation's own situation. Knowing what must be in the AI usage policy and who signs off on what.

Can decide and account for decisions on behalf of the organisation.

Measuring

How do you measure AI literacy?

In four steps, three of which are paperwork and one of which is the training itself. That sounds bureaucratic, but it is exactly the part that remains once someone asks about it.

01
Baseline measurement
A short assessment per employee shows what level already exists and where the real risks sit, often not where management expects them.
02
Role mapping
Based on the baseline measurement, you assign employees to a level. That differentiation is not a luxury: the law expects the level to fit the role.
03
Follow-up test
A knowledge check per participant after the training. Without a test, you have demonstrated attendance, not literacy.
04
Register
Everything recorded: who, when, which level, what result. Together with the AI usage policy, this is your evidence.

Included with every training

Three AI agents, ready to use.

Every training comes with three ready-made AI agents, chosen for your sector and set up around the way you work. They're live the moment the training ends, with no implementation project and no extra invoice.

01

Email Triage Agent

Sorts the inbox by topic and urgency, drafts replies to recurring questions, and hands anything that needs judgement to a person.

02

Knowledge Base Agent

Answers questions in plain language from your own documents, procedures and manuals, citing the source, so every answer can be checked.

03

Document Drafter

Produces first drafts of recurring documents, such as reports, letters and summaries, from data you already have, ready to review and send.

All agents run on European infrastructure, never use your data to train third-party models, and hand every decision that matters to a named person.

FAQ

AI literacy, made concrete.

What the term means, how it relates to a training programme, which level your team needs, and how you fund it.

AI literacy is the combination of knowledge, skills and understanding that lets someone use AI systems deliberately and recognise their risks, limits and consequences. It is not about programming: it is about knowing what a model can and cannot do, when the output needs to be checked, which data must never go into it, and when a human must make the decision. The term comes from Articles 3 and 4 of the EU AI Act.

AI literacy is the goal; AI training is one of the means. Article 4 mandates the result (employees who can use AI responsibly), not a specific course format. A demonstrably literate organisation typically combines training with a documented policy, role-based learning paths and a register showing who completed what.

That differs per role, and that differentiation is exactly what the law expects. We work with three levels: basic for all employees who use AI tools, advanced for people who apply AI in decision-making or customer contact, and accountable for leadership, procurement, legal affairs and privacy officers who decide which systems enter the organisation.

With a baseline measurement beforehand and a test afterwards, per role. The baseline measurement shows what level already exists and where the risks lie; the follow-up test delivers the evidence per participant. Both are recorded in the participant register, which together with the AI usage policy forms the compliance dossier.

Yes. It falls under professional development and sustainable employability, and so fits within the training budget and the personal development budget under the collective labour agreement (cao). For SMEs, the SLIM subsidy scheme additionally reimburses 60% of the training costs, up to a subsidy ceiling per application.