AI training for employees: what to teach, who goes first, and what to say alongside it

Most employees are using AI ahead of their training

Usage has outrun training almost everywhere. In BCG's survey of close to 12,000 people, three quarters of frontline employees already call themselves AI users, and only about a third feel properly upskilled. Pew found that of US workers who took any job training in the past year, about a quarter took training related to AI. Most people are teaching themselves, on whatever tools they found, with no shared rules about what data goes where.

That is the situation most training programs walk into, and it changes what the training is for. Few of the people you train will be meeting AI for the first time. Many will be wondering what their own use of it means for their job.

Training raises skill, and it can raise fear

Gallup tracked workers across four survey waves from 2023 to early 2026 and found that those who use AI daily or several times a week are more than twice as likely to fear their job will be eliminated within five years as those who use it a few times a month or year. Fluency shows people clearly what the tools can do, and what that might mean for them.

Pair every training block with an honest conversation about what the organization intends to do with the capability people are being asked to build.

Judgment Day, chapter 10

The same research found the link between AI use and that fear was weaker among workers who rated their manager highly. Training delivered by a learning platform, with no conversation around it, gets the skill and misses the trust. More on the evidence is in our answer to whether AI training reduces anxiety.

Train managers first

Employees whose manager actively supports their team's use of AI are 2.1 times as likely to use it a few times a week or more, and only 28% strongly agree their manager does (Gallup, 2025). Inside companies, the gap is visible from the top: in Gallup's 2026 survey of 102 chief HR officers, 57% said they now provide AI training for managers, and half said they were not confident their managers could guide employees on using AI at work.

So the program starts with managers, for two reasons. They need their own hands-on fluency before they can back anyone else's, and they need language for the question their teams will ask the moment training starts. The US Department of Labor's 2026 AI literacy framework makes the same call, listing manager upskilling as its own role. Our page on AI change management covers what managers can say when they cannot promise anyone their job.

The curriculum: what to teach, in what order, to whom

This is the sequence we run. It follows the five areas of the Department of Labor framework (understand, explore, direct, evaluate, use responsibly), taught hands-on, which the framework names as the most effective way to build literacy.

The weeks are a guide for a team of up to a few hundred people. The order matters more than the calendar.

Three rules hold the program together. Teach on real work, because people adopt what saves them time on Tuesday, not what impressed them in a demo. Give every role track enough hours to build a habit; BCG found regular use is sharply higher with at least five hours of training and access to in-person coaching. And keep practice going inside meetings the team already holds, so the skill does not fade the week after the course ends.

How to measure AI training

Completion rates tell you who sat through the course. They do not tell you whether anything changed. Measure four things instead, before the program and at the end of each 90 days:

• Usage. How often people use AI on real work, and for what.
• Confidence. Whether people feel prepared and comfortable using AI in their role.
• Manager support. Whether people say their manager actively supports their team's use of AI.
• Belief about their role. What people expect AI to mean for their own job.

Ask these directly and let people choose whether to answer. Report only for groups large enough that no individual can be identified, and never let one person's response reach their manager. Do not try to infer how people feel from their messages or their behavior. In the EU that is a prohibited practice under the AI Act, and everywhere else it is the fastest way to lose a team's trust. The AI readiness assessment covers this ground for a team in four minutes.

Is AI training legally required?

In the European Union, yes, in proportion to how you use AI. Article 4 of the AI Act has required providers and deployers of AI systems to take measures to support the AI literacy of their staff since February 2, 2025. An amendment in 2026 confirmed that no specific level is mandated, and the European Commission says no certificate is needed; an internal record of training is enough. National authorities began supervising the rule on August 2, 2026. It reaches organizations that deploy AI in the EU, which can include US companies with operations there.

In the United States there is no general requirement. The Department of Labor's AI literacy framework, published in February 2026, describes itself as voluntary guidance. Claims that the law requires a fixed AI course for every employee overstate both.