AI literacy training for employees: curriculum, hours and how to check it worked

AI literacy training for employees works best as a short core program, followed by practice on real tasks. Emeri suggests about six to ten hours across the first quarter. That time covers what AI tools do, how to use them on the job, how to check their output, and the rules that apply. Then check learning with a work sample, not a quiz alone.

There is no legal standard for the length or format of this training. For organizations under the EU AI Act, the European Commission says there is no single format, and no certificate is needed (European Commission). That leaves the design to you. This guide gives a starting outline.

What AI literacy means for an employee

The EU AI Act defines AI literacy as the skills, knowledge and understanding that let people make an informed deployment of AI systems. The definition also covers awareness of the opportunities, risks and possible harm of AI (AI Act Service Desk). For an employee, that becomes four practical questions:

• Which AI tools am I allowed to use, and for which tasks?
• How do I get a useful result from them?
• How do I know when the output is wrong?
• What data must I keep out of them, and who do I tell when something goes wrong?

A curriculum outline

The hours below are suggestions. Adjust them for the risk of the tools your staff use and the level they start at.

1. How AI tools work (45 minutes): what a language model does, why it can state false things with confidence, and what it cannot see.
2. Your organization’s tools and policy (45 minutes): the approved tools, the tasks they are approved for, and the data rules.
3. Asking well and giving context (90 minutes, hands-on): writing a clear request, supplying background and examples, and revising a weak result.
4. Checking the work (90 minutes, hands-on): finding invented facts, wrong figures and missing sources in AI output, using the team’s own documents.
5. Role tasks (two to three hours, hands-on): three or four real tasks from the person’s job, done with an AI tool and reviewed.
6. Risk, bias and incidents (45 minutes): where AI output can treat people unfairly, what counts as an incident, and how to report one.
7. Assessment (30 to 60 minutes): a work sample scored against a rubric.

Spend at least half the time on hands-on work. Employees learn to check AI output by checking AI output, not by watching a slide about it. Use the team’s own documents and tasks where the data rules allow. Emeri’s study guides cover the same skills, such as checking the work and keeping it safe.

Who needs more than the core

Not everyone needs the same depth. The Commission says organizations may differentiate training by staff knowledge, role and the risk of the AI systems involved (European Commission). A practical split:

• All staff who use AI tools: the core program above.
• People who review or approve AI-assisted work: extra time on checking, and on the review steps the team uses.
• People who oversee AI systems that make or support decisions about others, such as hiring tools: training on that specific system, its limits and when to override it.
• Managers: how to set expectations for AI use, review AI-assisted work, and handle an incident.

Formats that fit

• Self-paced e-learning: good for the concepts and the policy module. It scales well and leaves a completion record.
• Live workshops: best for the hands-on modules. A facilitator can see where people get stuck and correct it.
• Team practice sessions: a team works through its own tasks together, which ties the training to the job.
• Office hours or an internal help channel: a place to bring real problems after the training ends.
• Short refreshers: a 20-minute update when tools or policy change.

Off-the-shelf courses can cover the concept modules. Two examples: Elements of AI, a free online course from MinnaLearn and the University of Helsinki (Elements of AI). And Google AI Essentials, which Google lists at under five hours over five modules. Google lists it at $49 a month on Coursera in the US and Canada after a seven-day trial, as listed in October 2026 (Google). Neither covers your own tools, policy or tasks. You still need to write those modules.

The Commission keeps a living repository of AI literacy practices from companies and public bodies. It shows how other organizations structure training. The Commission notes that copying a practice from the repository does not by itself mean an organization complies with the AI Act (European Commission).

How to check that learning happened

The Kirkpatrick Model is a common way to evaluate training. It has four levels: reaction, learning, behavior on the job, and results for the organization (Kirkpatrick Partners). Reaction is the easiest level to measure and says the least about skill. Aim for the second and third.

• Reaction: a two-question survey at the end. Was it relevant to your work? What was missing? Useful for fixing the course, not for judging skill.
• Learning: a work sample at the end of the program. Give a real task, collect the output and the drafts, and score them against a rubric.
• Behavior: 30 to 60 days later, review a sample of real AI-assisted work from each team. Look for the checks the training taught.
• Results: track a small number of measures you already collect, such as error rates in reviewed work or time to complete a task type.

A multiple-choice quiz can confirm that people read the policy. It cannot show that they can check an AI draft. Use quizzes for the rules, and work samples for the skills. The AI literacy assessment page explains the difference in more detail.

Keep a record

Record who completed which modules, when, and the result of any assessment. The Commission says organizations under the AI Act can keep an internal record of training and other guidance (European Commission). A simple log works. The guide on EU AI Act Article 4 covers the legal duty, and documenting AI literacy training covers the record itself.

A rollout plan

1. Publish the AI use policy before the training starts. Emeri’s team AI use policy template is a starting point.
2. Run a baseline work sample with a pilot group of 10 to 20 people.
3. Deliver the core program to the pilot group over four to six weeks.
4. Repeat the work sample, and compare the scores.
5. Fix the modules where scores did not move, then roll out to everyone.
6. Review a sample of real work after 60 days.

For a baseline before training, staff can take the free check, which takes about 15 minutes. For a list of topics to cover in a shorter awareness session, see AI awareness training topics.