Which AI literacy measures can a company take besides a course?

The useful question is not “which course should we buy?” but “what does each group need to understand to use this AI well?” Format follows the system, the person’s role and the possible effects.

There is no single mandatory course

The Commission says there is no one-size-fits-all solution. Organisations should consider their systems, prior knowledge, experience, use context and the people on whom the use may have an impact.

Measures for lower-impact uses

A concise use rule, examples of data that must not be entered, and a human-review step may be more useful than a generic two-hour training session.

Measures for specific functions

HR may need bias and decision-review criteria. Sales may need limits for proposal drafting. Support may need escalation rules. The content changes because the decision changes.

Measures for sensitive systems

When an output affects people or significant decisions, a targeted session, supervised practice, operating instructions and periodic incident review may all be appropriate.

• A concise policy or acceptable-use rule.
• Guidance for a specific tool or task.
• Role-based briefings using real cases.
• Supervised onboarding before independent use.
• A question channel, incident reviews and reminders.
• Structured training where the context justifies it.

Proportionality should also be documented

Choosing a measure is not enough. Record why it fits that group and use. The Commission’s examples are helpful, but following one does not by itself create a presumption of compliance.