AI literacy for SMEs: a practical 30-day plan under Article 4 of the EU AI Act

Article 4 AI literacy is enforceable since August 2026. A practical 30-day plan for SMEs: no certificates, but documented, role-based training.

Since 2 August 2026, national market surveillance authorities can supervise and enforce Article 4 of the EU AI Act. The obligation itself is older: it has applied since 2 February 2025. If your people use ChatGPT, Copilot, Gemini or any AI feature inside your software, AI literacy is now a concrete task for management, not a theoretical one.

What the rule really asks

According to the European Commission's Q&A, providers and deployers of AI systems must take measures to support AI literacy of their staff and of other people who operate AI on their behalf, such as contractors. There is no mandatory training format, no certificate and no test of individual knowledge. The Digital Omnibus on AI, in force since 27 July 2026, kept the duty but confirmed that nobody has to guarantee a specific level of literacy for every individual.

What the Commission does say: measures should fit your context. They should consider your role (provider or deployer), the risks of the tools you use, and your people's existing knowledge and experience. Simply handing out the tool's instructions is described as often ineffective. And it is sensible to keep an internal record of trainings and initiatives. The Commission also notes that sanctions are decided nationally, proportionately, and that penalties are more likely where an incident is linked to poor training or guidance.

A 30-day plan for a small company

Week 1: map who uses what

List every AI tool in use, official or not, who uses it, for which tasks and with what data. Include AI features hidden in tools you already pay for. Most companies discover more usage than they expected, including personal accounts used for work.

Week 2: write a one-page usage policy

Say which tools are approved, which data must never be pasted into them (customer personal data, contracts, credentials), who reviews AI output before it reaches a customer, and who to ask when in doubt. One page that people read beats twenty pages that nobody opens.

Week 3: train by role, not by headcount

Different groups need different content. Sales and marketing need to check facts, avoid leaking client data and disclose AI-generated content where required. Administration needs data-protection basics. Managers need to understand risk, accountability and when a use case may become high-risk. Technical staff may still need the legal and ethical side. Short practical sessions with real examples from your own work are more effective than a generic video.

Week 4: document and set a review date

Keep a simple register: date, topic, who attended, materials used. Add the policy version and a date to review it, for example every six months. This is the evidence you can show if an authority, a client or an auditor asks what you did.

Where the rule gets stricter

Article 4 is the baseline. If you deploy a high-risk system, for example for recruitment or credit decisions, Article 26 adds a specific duty to have staff competent to oversee it. Under the Omnibus the stand-alone high-risk rules now start on 2 December 2027, so there is time, but the training habit is easier to build now.

Common mistakes

  • Treating it as a one-off event instead of a routine that follows new tools and new staff.
  • Training only the IT team while sales and admin use AI every day.
  • Buying certificates without changing any daily behaviour.
  • Forgetting freelancers and agencies who work with AI on your behalf.

At Vivig we build this kind of foundation in our AI Evolution Program: a map of processes, a usage policy and documented, role-based training for each department. If you want to start with a short conversation about your current situation, you can get in touch. This article is general information, not legal advice.

Want to bring these ideas into your company?

Let's talk →