More AI training can mean less AI use. Feature courses without changed expectations teach people that AI is optional. Train on their own tasks, in the manager's meeting rhythm, not in the learning platform.
Short answer: AI upskilling works when it starts from employees' own tasks, happens in several steps over time, is led by the line manager, and is followed up by looking at what actually changed in the work. A single course builds knowledge, but rarely changes behaviour.
The problem with most AI training
Search for "AI training" and you mostly find courses: online, free, certificates, university modules. Fine for individuals who want to learn AI. For an employer who wants AI to become part of the job, a course catalogue is rarely the answer.
The typical format: an introduction to generative AI, a session on writing prompts, a few examples from other industries. Everyone gets a certificate. Then people go back to their desks, where nothing has changed. Same goals, same routines, same expectations from the manager.
The message that lands is not "this is how we work now" but "here is something to try if you have time".
What research says about why AI goes unused
Prosci's study of 1,107 professionals found user proficiency is the biggest single AI implementation challenge, at 38% of reported difficulties. It breaks down into the learning curve (22%), prompting difficulties (11%) and inadequate training (6%) (Prosci).
Notice the last number. The main problem is not a shortage of courses. It is the learning curve inside people's own work. You don't solve that with more courses. You solve it with support where the work happens.
The fear nobody talks about
Many employees wonder, often quietly, whether AI will take their job. Training that is only about features and efficiency can amplify that fear, especially if leadership is talking about savings at the same time.
Talk about it openly. Say what you know and what you don't. Explain which tasks you hope AI will take over, and what that means for roles. Few people believe vague promises that "nobody will lose their job". Concrete reskilling plans are more credible.
How to build AI skills that last
1. Start with the task, not the tool
Ask the team which task they would most like to never do again this week. Start there. It builds motivation from the beginning.
2. Train teams that work together
A team learning together, on its own documents and cases, builds shared ways of working. Individuals taking courses alone only create isolated enthusiasts.
3. Break it into short steps
A first session of 60 to 90 minutes, then two to four weeks of everyday use, then follow-up. The follow-up is where the real learning happens.
4. Let the manager own it
Managers set expectations. If they use AI themselves, ask about it in the weekly meeting and highlight good examples, it becomes part of the job. Train managers first.
5. Teach checking, not just using
People freeze when they don't know when an AI answer can be trusted. Give them a simple routine for verifying outputs. That builds the confidence to use the tool.
6. Make it safe to fail
If early attempts are judged as finished work, people stop experimenting. Say clearly that early attempts are drafts.
7. Link it to the EU AI Act
Since 2 February 2025, Article 4 of the EU AI Act has required organisations using AI systems to take measures to support the development of AI literacy among staff, taking role and context into account (EUR-Lex). Role-based training on real tasks is better learning, and a stronger answer to the legal requirement than a generic course.
A simple plan for one team
| Week | Activity | |---|---| | 1 | Manager and team pick three tasks. Short intro and rules. | | 1–2 | Hands-on session: everyone does the three tasks with AI, in pairs. | | 2–4 | Everyday use. One item on the weekly meeting: what worked, what didn't? | | 4 | Follow-up: what has changed? Which tasks next? | | 8 | New round with the next tasks. |
People are the bigger part
At Lyon AM our view is that roughly 70% of a successful AI programme is about people, leadership and ways of working. Training is an important part of that, but only when it is tied to changed expectations and ways of working.
Frequently asked questions
How much AI training do employees need? It depends on role and usage. More important than hours is that training uses real tasks and is followed up over time.
Should AI training be mandatory? A baseline on rules and risks should apply to everyone using AI at work. The practical part works best when tied to the team's own tasks.
How do you reduce fear that AI will take jobs? Talk about it openly, be specific about which tasks will change, and show a reskilling plan. Vague reassurance doesn't work.
Who should lead AI training? Ideally the line manager together with internal role models. External trainers can provide the foundation, but the change happens in the team.
Next step
Want to know where the skills gap really is? The Lyon AM AI diagnostic shows both usage and foundations. Contact us to build AI skills that show up in the work.
