Half Your Employees Had AI Sit Through Their Training. You Might Never Know.

Half of your employees have used AI to complete mandatory training. Not to help them study it. To sit through it instead of them.

A Moodle survey of American workers found that 52 percent have used AI tools to get through required training: 21 percent lean on AI for the hard questions, 19 percent let it write their answers, and 12 percent hand over the entire course and go do something else. Another 46 percent speed up the videos or let them play in the background. The training happened. The person did not.

What Is Actually Happening When AI Takes the Training For You

This is not a small workaround by a handful of disengaged employees. It is half your workforce, and it is not really about laziness. Two-thirds of American workers report burnout, and the number climbs past 80 percent for employees under 35. When a mandatory course lands in someone’s inbox on top of an already full week, letting AI click through it is the fastest way to make the notification disappear.

The training platform records a completion. The dashboard turns green. Somewhere in a compliance report, the box is checked. But the thing the training was supposed to build, whether that is a new skill, a policy understanding, or a safety habit, never landed anywhere. You have a record of learning with no learning inside it.

Why the Old Model Was Already Vulnerable

Mandatory training has run on the same assumption for twenty years: if someone clicks through the slides and passes the quiz, they know the material. AI did not break that assumption. It exposed how thin it always was. A course built to be watched, not practiced, was never going to survive a tool that can watch things for you.

This is the same pattern that shows up everywhere AI meets an organization built on checking boxes instead of building capability. The visible system says everything is fine: completions logged, certificates issued, audit trail clean. The invisible system, the actual readiness of your people to do the thing the training was for, tells a different story. Leaders who only look at the visible system find out about the gap at the worst possible moment: an audit, an incident, a customer complaint that traces back to something “everyone was trained on.”

Does Mandatory Training Even Work If AI Can Do It For You?

Only if you change what you are actually measuring. Completion was always a weak proxy for competence, and AI just made the proxy free to fake. The fix is not banning AI from training, which is a fight you will not win and do not need to. The fix is building training that has no shortcut because the value sits in something AI cannot do for your employee: a real decision, made in front of someone who is paying attention.

A few shifts worth making before your next mandatory training cycle:

• Replace one passive module with one live conversation. A 15 minute check-in where someone applies the material to their actual job teaches more than a 45 minute video, and it cannot be delegated to a bot.
• Ask people to teach it back, in their own words, to a real person. AI can summarize a policy. It is much worse at standing in for your employee’s judgment when a manager asks, “what would you do if this happened on your floor tomorrow?”
• Track application, not completion. A completion rate tells you who clicked through. A short follow-up, a spot check, a manager conversation two weeks later tells you who actually changed how they work.
• Say the quiet part out loud. Tell your team you know AI can finish this for them, and tell them why that is a loss for them specifically, not just a compliance risk for you. People route around rules they do not understand. They are much slower to route around a reason that makes sense.

What This Costs You Beyond the Training Itself

The completion report is not the only thing that goes quiet. Once employees learn that gaming one mandatory course carries no real cost, that lesson generalizes. It tells them the organization cares about the record of learning more than the learning. That is a dangerous thing to teach a workforce, because the next thing they apply it to might not be a video about workplace safety. It might be a quarterly report, a customer commitment, or a task they tell you is “done” because the box got checked.

Leaders often treat this as an L&D problem to hand off to whoever manages the training platform. It is not. It is a trust and attention problem, and those sit with leadership, not with the software. A manager who never asks what a course actually covered is teaching their team, just as clearly as any policy document, that the answer does not matter. A manager who asks one specific question in the next one-on-one, tied to the training their report just “completed,” teaches the opposite lesson in under a minute.

The Part AI Cannot Sit In For

Here is the tension in that Moodle number that should bother every leader reading it: your people found the most human response available to an inhuman amount of required learning. They were exhausted, so they used a tool to protect their time. That is not a discipline problem. That is a design problem, and it is fixable.

I have spent years standing in front of rooms telling them AI can do more than they think, and then spending the rest of the session telling them not to let it do the part that actually matters. Training is one of those parts. The value was never in the video playing. It was in a person wrestling with a real scenario, out loud, in front of someone who could tell whether it landed. AI can prep the material, personalize the path, and flag who is behind. It cannot have the conversation that makes the learning stick. That part still needs a human in the room, paying attention, making the invisible work of understanding visible again.

If your last training rollout produced a clean completion report and you still are not sure your people could handle the real version of what it covered, that gap is the actual story, not the dashboard. Training that survives contact with AI does not look like a course library. It looks like practice, feedback, and a real person checking whether it worked.

That is the training worth building next. If your team’s last rollout says “complete” everywhere and you still are not confident it changed anything, let’s build the version that does.