How to prove AI literacy to a supervisor under Article 4

Article 4 of the EU AI Act requires providers and deployers to take measures, to their best extent, to ensure a sufficient level of AI literacy. The European Commission does not provide a fixed checklist or mandatory certificate. That flexibility is useful, but it also leaves organisations asking a practical question: what should we show if a supervisor asks?

The strongest route is an evidence pack that fits your actual AI use. Not a folder of disconnected course certificates, but a clear chain from scope to roles, from roles to learning goals, from learning goals to training and assessment, and from results to management decisions.

What will a supervisor likely want to understand?

A supervisor will mainly want to see whether your approach is logical. The question is not: "did everyone click through a course?" The question is: "do the measures fit the AI systems, the risks and the people working with them?"

A strong Article 4 evidence pack therefore includes:

• A current overview of AI tools and AI systems
• A role matrix: who uses, manages, develops, assesses or decides with AI
• Learning goals per role and risk context
• Training, work instructions and practical guidance
• Attendance and assessment records
• Certificates as supporting evidence
• Management reporting with open gaps and follow-up

This is where many organisations get stuck. They delivered awareness sessions, but cannot explain why those sessions are sufficient for HR, legal, customer contact, IT or data science teams.

The evidence chain in five steps

Step 1: start with AI use, not training

Start with AI use in the organisation. Which generative AI tools are allowed? Which SaaS applications include AI functionality? Where is AI used for screening, advice, classification, customer contact or decision preparation?

Without this scope, AI literacy becomes generic. Generic training is weak evidence because the AI Act points to context: technical knowledge, experience, education, the use context and the people affected by the system.

Step 2: connect systems to roles

Not everyone needs the same knowledge. A recruiter using AI in selection needs to recognise different risks than a marketer generating copy or a data engineer monitoring models.

Make clear per role:

• Which AI systems or tools are used
• Which decisions or outputs are involved
• Which risks the role must recognise
• When human review or escalation is required
• Which documentation or logging is expected

Step 3: translate roles into learning goals

A learning goal is stronger than a course name. For example: "HR staff can explain when AI use must be made transparent to candidates" is more testable than "HR follows AI awareness".

Good learning goals include behaviour. Think of checking output, avoiding sensitive input, recognising bias signals, validating sources, escalating uncertainty and documenting AI use.

Step 4: record training and assessment

Use training records to capture attendance, score, certificate, date, role and validity. Also record exceptions: who has not completed the path, who scored too low, and which remediation is running?

You can start with the Article 4 Evidence Dossier Checklist and then use the AI Training Records template for employee-level records. For larger teams, a platform such as LearnWize is more practical because assessment, learning path, certificate and team reporting stay together.

Step 5: make management accountable

AI literacy is not an HR project that ends after an e-learning module. Management should see:

• Which roles have completed the required learning
• Which teams still have risk exposure
• Which AI systems require additional training
• Which incidents or near misses create new learning goals
• When policies or onboarding are updated

That keeps the evidence pack alive. This matters because AI literacy is an ongoing process, not an annual checkbox.

Certificate: useful, but not sufficient

An online AI literacy certificate is useful evidence if it is concrete. It should show who completed what, when, with which result and for which role. But a certificate without context does not prove that the organisation ensured a sufficient level of AI literacy.

Use certificates as part of the evidence pack, not as the evidence pack itself.

Where LearnWize fits

When you move from documentation to team execution, LearnWize is most useful for measuring and recording AI literacy at team level. Use LearnWize for:

• Readiness assessment per team
• Role-based modules
• Assessment outcomes
• Certificates as supporting evidence
• Team dashboard and progress reporting

See the pillar page How to prove AI literacy to a supervisor for the full evidence approach.