AI literacy for employees: why microtraining works

Employees who use AI in their work do not need to become AI specialists.

They do need to understand enough about what AI can do, where the risks lie and how to use it responsibly.

The EU AI Act requires organisations that provide or use AI systems to take measures relating to AI literacy. What Article 4 asks in concrete terms is set out here. The Digital Omnibus changed that obligation but did not remove it.

The practical question is therefore: how do you effectively transfer a limited amount of important knowledge to a large group of employees?

Microtraining suits that kind of knowledge particularly well.

Microtraining: small amounts of knowledge you can use straight away

For many employees, AI literacy comes down to concrete things:

Microtraining splits that knowledge into short, focused parts.

Research into microlearning, the educational term for learning in small units, supports this approach. A systematic review of forty studies found positive effects on knowledge acquisition, memory, skills and motivation, among other things.

That makes this format interesting when AI is only one part of daily work.

Reading alone is not the same as learning

Putting an AI policy on the intranet or asking employees to read a set of instructions is easy.

But that still does not tell you what someone has understood and remembered.

The European Commission also warns that relying only on instructions for use, or asking staff to read them, may be ineffective.

AI literacy is therefore more than a tick in a box.

A short training course adds something: information is given structure, employees practise with it and mistakes can be corrected straight away.

Practice questions make the difference

Research into retrieval practice shows that actively recalling knowledge can work better than reading the same information again.

A meta-analysis examined 192 effect sizes from 122 experiments and 67 publications, covering 10,396 participants in total. The average transfer effect was d = 0.40.

But not every quiz works.

The effect is stronger when practice questions resemble what someone will actually need to be able to do later and when there is enough practice. Random factual questions therefore add little.

Take this example:

An employee copies customer data into a public AI tool to have it summarised. What is the risk here and what should they check first?

A question like that gets someone to apply knowledge to a recognisable work situation.

That is different from simply checking whether a rule has been memorised.

Why have a test at all?

A final test is not mandatory under the EU AI Act. Nor does a certificate prove that an organisation is automatically compliant.

A test does have value.

You check whether the most important knowledge has been understood, and you can record who took the training, what was covered and what result was achieved.

Whether that knowledge then shows up in daily work is a next step. That is the subject of moving from AI training to behaviour in the workplace.

The logical sequence then becomes:

transfer knowledge → practise → test → record

This is already happening in practice

This approach is not just theory.

According to the European Commission's repository, Hexagon uses short basic training courses of roughly fifteen minutes. Completion is tracked and employees receive certificates.

Alteryx offers AI knowledge across the whole organisation and adds role-based training on top of that. Employees with access to internal AI models must complete the relevant training and pass an assessment.

Any organisation that wants to introduce AI literacy organisation-wide can then do so step by step. Our guide to AI literacy sets that out in five phases.

AI literacy does not have to become a large training project

Not every employee needs to spend days on a course. A short AI training course can be valuable precisely when the content is limited to what someone needs for their own work.

That is also the principle behind the AI literacy training courses from AIAdopt.

The EU AI Act modules consist of six lessons, with two practice questions after each lesson. The training ends with a test of ten questions. On passing, the participant receives a certificate listing the learning outcomes covered and the articles addressed.

The law does not prescribe this format.

But if, as an organisation, you are going to make an effort on AI literacy anyway, it makes sense to choose an approach that transfers knowledge, has employees practise with it and checks how much of it has been understood.

Do you have a question about the approach in your organisation? Get in touch.

Sources

AI Literacy - Questions & Answers, European Commission

Repository of AI literacy practices, European Commission

Microlearning beyond boundaries: A systematic review and a novel framework for improving learning outcomes, Heliyon

Transfer of test-enhanced learning: Meta-Analytic Review and Synthesis, Psychological Bulletin