How to Run AI Training When Half the Room Is Advanced and Half Is a Beginner

The most common reason AI training fails in mid-market organizations has nothing to do with the material.

It has to do with who is in the room at the same time.

A senior analyst who has been using ChatGPT daily for eighteen months. A plant supervisor who has never opened the tool. Both sitting in the same session, expected to get the same value from the same content.

The advanced user checks out in the first ten minutes. The beginner falls behind and stops asking questions. By the end, neither group has what they need.

This is not a training design problem that can be solved with better slides. It is a structural problem that most organizations do not see until they are already running the session.

Why the Mixed Room Happens

The instinct behind bringing everyone together is reasonable. AI is an organizational initiative. The rollout should feel unified. Running separate sessions takes more time and more coordination.

So the session gets designed for the middle, not too basic, not too advanced, and it ends up serving neither end of the room well.

Advanced users in these sessions describe a familiar experience: they already know everything covered in the first half, and by the time the session reaches anything new, the beginners are lost and the conversation has stalled.

Beginners describe a different version: they are managing the anxiety of not knowing something everyone else seems to understand, they do not want to look uninformed by asking foundational questions, and they leave with the surface impression that AI is useful without any clear idea of how to use it on their actual work.

The session ends. The split remains.

What Actually Works for Mixed Rooms

The organizations that get consistent results from AI training across different skill levels have usually done one of two things.

The first is separate tracks. Not two entirely different programs, but a deliberate split in how the session is structured. Advanced users get workflow depth: here is how to use AI on the specific reporting, analysis, or communication work you already own. Beginners get foundation plus one workflow: here is what the tool does, and here is exactly how to try it on something you will do next week.

This requires more coordination upfront. It produces meaningfully better results because each group leaves with something they can actually use.

The second approach, when separate tracks are not feasible, is to anchor the entire session to work the team already does rather than to the tool's capabilities. Starting from a report everyone builds, a document everyone creates, or a workflow everyone runs removes the competence gap from the center of the room. The advanced user and the beginner are both looking at familiar work. The question becomes how AI fits into that work, not whether the person understands AI.

In a recent workshop with a manufacturing team that used this approach, 95% of participants said they planned to apply what they learned within two weeks. For comparison, that number typically sits between 40 and 60% after standard AI training. The mixed room did not disappear. The starting point changed what the room was working on together.

The Super-User Problem

There is a related issue worth naming.

Most organizations have a small number of people who are genuinely advanced AI users. These individuals are often identified as internal champions and asked to help train their colleagues.

The problem is that expertise in using a tool does not transfer easily into expertise in teaching it. Advanced users have often forgotten what it felt like not to know how to use it. They skip past foundational questions that beginners need answered. They use vocabulary that is not yet shared. They demonstrate things at a speed that feels natural to them and opaque to everyone else.

This is not a criticism of the super-user. It is a structural issue. The skill that makes someone effective at using AI is different from the skill that makes someone effective at helping others use it.

When organizations rely primarily on super-users to train the rest of the organization, they often produce a second round of uneven adoption, just in a different configuration than the one they started with.

What to Do Before the Next Training

Before designing another AI training session, answer three questions.

Who is actually in the room, and what does their current experience with the tool look like? Not the organizational average, the actual range. If the range is wide, the session probably needs to be split or anchored differently.

What specific work do the people in the room already do that AI could support? The answer to this question is the starting point for the session, not a closing exercise.

What does success look like two weeks after the training? If the answer is not specific, the session is not yet designed to produce it.

If your organization has run training and adoption is still uneven, the problem is usually upstream of the content. The AI Operations Review is a diagnostic conversation that identifies whether the constraint is the training design, the workflow structure, or something else, before you run the next session.