How do you train a team with mixed AI skill levels?
Assess starting points, group people by role and confidence, set a shared baseline for safe use, and let confident users support peers without taking over.
Why one-size sessions lose both ends
When a single session tries to serve everyone, it suits nobody well. Beginners are asked to try things in front of colleagues who seem fluent, which can feel exposing. Confident daily users sit through basics they already know and switch off. The middle of the group may do fine, but the people at either end leave with little to show for it.
Find out where people are first
Before planning content, assess starting points. Keep it light and non-judgemental so nobody feels tested.
Short self-assessment: Ask which tools people have tried, what they use them for and how confident they feel, using plain options rather than technical terms.
Task-based check: Ask each person to describe one task where they have used AI or would like to, which shows practical understanding better than a self-rating.
Manager input: Ask line managers which tasks in each role are most likely to benefit, so starting points connect to real work.
Make clear that the aim is to plan useful sessions, not to rank people. Some of the least experienced users will have thought hardest about the risks.
Group by role and confidence
Grouping by confidence alone can create a beginners' group that feels singled out. Combining role and confidence works better.
Role-based groups: Put people with similar tasks together so examples are relevant, for example admin staff working on scheduling and correspondence, or finance staff working on reconciliation notes.
Tracks within sessions: Offer a core exercise for everyone and an extension task for those ready to go further.
Optional deeper sessions: Run separate, opt-in sessions for confident users on more advanced work, such as building reusable instructions for repeated tasks.
What each role actually needs to learn is covered in what AI skills a team actually needs.
Set a shared baseline for safe use
Whatever their starting point, everyone needs the same foundation. Confident users are sometimes the most likely to have picked up risky habits, such as pasting client information into personal accounts.
Approved tools: Which tools people may use for work, and which they may not.
Data rules: What information must never go into an AI tool.
Checking output: The expectation that every AI-assisted piece of work is reviewed before it goes anywhere.
Where to ask: Who to contact with questions or concerns.
Data protection duties differ between organisations, so take advice on your own situation when setting these rules.
Use confident users as peer support
Confident users can help others, but they need a defined role so they do not dominate.
Pair rather than present: Pair a confident user with a newer one on a shared task, instead of asking them to demonstrate to the whole room.
Ask them to share failures: Invite confident users to talk about what went wrong for them, which makes it easier for beginners to have a go.
Set limits: Ask confident users to guide with questions and let their partner do the work, rather than taking over the keyboard.
If peer support proves valuable, you may want to formalise it, as discussed in whether you need AI champions in your team.
Regroup as people progress
Starting points change once people begin practising. Recheck confidence after a few sessions and regroup where it makes sense, so nobody is held back or rushed.
More questions