How Do You Train People Who Learn at Different Speeds?
Last Updated: September 2026
Mixed-ability AI training is what you run when one team carries several levels of AI skill at the same time. A few people already build their own workflows. Most have tried a tool, got something useful, and never repeated it. A smaller group has barely started. The lesson pitched at the middle of that spread runs too slow for the front and too fast for the back, so one plan loses the two groups you can least afford to lose.
AI Smart Ventures has guided growing businesses through AI adoption long enough to watch how wide that spread runs inside a single team. The gap is rarely about job title or age. It comes down to how many hours someone has already spent with the tools on their own time, and nobody in the room knows that number.
Ignore the spread and the bill arrives twice. Your strongest users stop turning up, because they knew the material before you booked the room. The people at the back learn that the safest move is to sit quiet. Six months on, you have one confident builder, a polite majority, and no shared standard.
Key Takeaways
- Most of your team sits in the middle: Section’s AI Proficiency Report put 73.5% of 5,026 US knowledge workers at the Experimenter level in June 2026, using AI often with no set routine behind it.
- Both ends are small, and both are easy to lose: the same study put 20.9% at Novice and 5.5% at Practitioner or Expert, so the average lesson serves neither.
- Most people are teaching themselves: 62% of workers expect to learn AI on their own and 53% from video tutorials, JFF’s worker survey found, against 9% who ask their employer.
- Sort quietly, then teach one task at three depths: same job, three finish lines, so nobody waits and nobody drowns.
Those numbers say something awkward. The spread is not proof that your training failed. It is what happens when people learn on their own time at their own pace. So the goal is not to level everyone before you teach. It is to design a session the gap cannot break.
Why does teaching to the middle lose both ends?
Because the middle is an average, and almost nobody sits on it. Aim an hour at the person who has used AI a handful of times, and the front row hears what it knows while the back row misses step one and spends the hour pretending. Both ends draw the same view: the session was not meant for them. The middle does learn something, and that is what hides the failure.
The two ends then fail at different speeds. People who are ahead get restless first. Thomson Reuters’ Future of Professionals 2026, published on 22 June 2026 from more than 1,800 professionals in 62 countries, found 91% had felt the gap between what AI was meant to give them and what it gave. Almost three in ten mid-career workers in it would change jobs within two years if it stayed open. The back group fails quietly, and you find out later.
What are the three speeds inside one team?
Three, in most teams: a small group already building for others, a large middle that uses AI without a system, and a group that has barely begun. Section’s AI Proficiency Report, out in June 2026 after testing 5,026 US knowledge workers on real skill rather than self-rating, sorts people into four named levels: Novice, Experimenter, Practitioner and Expert. It found 20.9% at Novice, 73.5% at Experimenter, and 5.5% at Practitioner or Expert combined.
Hold those shares against your own team and the shape turns useful. The middle is not idle. It has real use with no routine underneath, so whatever it gains fades the week things get busy. The front group has crossed into building things other people run, which is a different job. The back group is rarely against any of this. It is behind on hours, and short of nerve about fumbling in front of people who are not.
| Speed | What they already do | What they need next |
|---|---|---|
| Ahead | Build workflows others use | A job to own, not more course material |
| Middle | Use AI often, start fresh each time | One task turned into a routine |
| Not started | Watch, and wait to be asked | A first task with the steps written down |
How do you find out where each person sits?
Watch one real task instead of running a quiz. Give the whole team the same short piece of work they already do, half an hour, any tool they like, then collect the output and the route taken to it. The route tells you more. Who opened a blank chat and typed one line, who pasted last month’s example in first, who checked the answer against a source, and who quietly asked someone before starting.
Asking people to rate themselves will not get you there, because self-taught skill is patchy in ways the person cannot see. JFF’s survey of 3,020 US residents, fielded between 28 November and 8 December 2025, found 62% expect to build AI skills on their own and 53% by watching video tutorials, while only 9% would ask their employer for help. In that survey, 17% felt not prepared at all to use AI at work.
How do you run one task at three depths?
Pick one job the whole team already does, then set three finish lines on it. Everyone starts from the same brief, so nobody gets sorted into a tier in public. The back group finishes the task once, with the steps written out beside them. The middle group finishes it, then writes the version that runs again next week. The front group builds the checked version the team will use, and that becomes the standard.
Depth does the sorting, so you never have to. Pair someone from the front with someone from the back for the first pass only, then let them split, because a lasting pair turns into one person doing the work. Leave the middle working alone, since that group learns by finishing something. Generic AI consultants tend to ship the same deck into every room, because a deck travels well and a sorted task does not. Run it by depth and you get capability building across the whole spread instead.
That is the design work, and it repays doing well. AI Training builds AI literacy around the jobs your team already runs, so the fast and the slow leave one session with different, useful things.
What keeps the fast group from drifting off?
Give them something to own rather than something to sit through. People at Practitioner level are not short of material, and another course will not hold them. Hand over one thing the team will really run: the checked prompt library for a repeat job, the review standard for AI output, the note of what the tool got wrong last month. Ownership holds attention where a syllabus never will, and turns one person’s skill into the team’s.
There is a second reason to look after that group. The New York Fed’s Liberty Street Economics, writing on 14 April 2026 from the November 2025 Survey of Consumer Expectations, reported that 39% had used AI tools at work, while just 15.9% said their employer offers AI training. Most of the skill in your team was built off the clock. AI Smart Ventures observes that people who built it that way notice when nobody asks them to use it.
Frequently Asked Questions
Should you split a mixed team into separate sessions?
Usually not, at least not at the start. Splitting by level creates a visible ranking, doubles what you have to deliver, and cuts off the fastest route the back group has, which is watching someone work. Keep one room and split the depth of the task instead, since AI enablement moves faster when the levels can watch each other work. Separate sessions earn their place later, when your advanced group is building agents, and the rest are not.
Which AI jobs suit each speed on your team?
Match the job to the hours behind the person. Someone starting out should sum up a long document, draft a reply for review, or tidy a messy spreadsheet, all frequent work where a mistake costs little. The middle group turns one of those into a routine with a saved brief and a check step. Your advanced people build the shared version and hand it over with the failure modes written down.
How do you start using AI when you are the slowest?
Pick one thing you already do every week and do that thing with AI three times. Not a new task, and not a big one. Tell the tool what the job is, what good looks like, and what to avoid, then read the answer against something you know is right. The third attempt is where it starts to feel normal. Speed comes from repeat runs on usual work.
Can someone sit out while the rest of the team learns?
For a while, yes, and it is often the right call. Forcing someone who is not ready into the room produces a quiet body in a chair rather than a user. Ask what pace would work instead, and offer a smaller first task with a partner. Where AI already sits inside software the job needs, sitting out stops being an option, so say that early and give people a runway.
Where does automation fit for people at different speeds?
Automation means having software finish a set step with nobody touching it, such as filing an attachment or moving a record between systems. It suits the front group, because writing one takes judgment about what should never run alone. Everyone else is better served by AI help, where a person still reviews each output. Automate a task only once someone has run it by hand enough to know how it fails.
Who should pair with whom when abilities differ?
Pair the nearly-there with the just-started, not the expert with the beginner. Someone one step ahead remembers what confused them and explains it in plain words, while an expert has forgotten the difficulty exists and moves too fast. Keep the pair to a single task, then break it. Long-standing pairs turn into one person doing the work and the other watching.
How long before the back group reaches the middle?
Give it a few weeks of real use, not a training day. Someone doing one usual task with AI two or three times a week tends to stop needing the written steps within a month, and stops asking whether the output is good enough soon after. The jump from there to building routines takes longer and needs a reason to exist. Measure the shift in the work, not the hours booked.
Does a skills check embarrass the people at the back?
Only if you run it as a test with a score. Watching one normal task, with no grade attached and results reported at team level, reads as work rather than an exam. Say beforehand what you are looking at, which is where the tool helps and where it wastes time, and say what the result will never be used for. An AI readiness check done that way draws honest answers.
What if the manager is the slowest person in the room?
That is common, and it is fixable once the manager stops hiding it. Managers get less spare time to play with the tools than the people they lead, so they fall behind on hours rather than ability. The move that works is to take the same first task as everyone else and say out loud that it is a first attempt. A team learns more from that than from any AI adoption policy.
Where do we start when the team is spread across levels?
Start by picking the one task everybody does, then watch how five people actually do it. That single hour usually shows you the shape of your spread and where the practical AI gains sit, before you commit to any AI upskilling plan. Design the three depths after that, never before. Schedule a consultation with AI Smart Ventures to sort the spread and set the task worth training on.
Executive Summary
Mixed ability AI training starts from an awkward fact: one team holds several levels of skill at once, and the average lesson fits none of them. Recent testing of 5,026 knowledge workers put 73.5% in the middle band, 20.9% at novice level, and 5.5% at practitioner or above. Most of that skill was self-taught. The fix is not to level everyone first. Pick one task the team shares, set three finish lines on it, pair the nearly-there with the just-started, and give your advanced group something to own.
What Should You Do Next?
This week, name the one task your whole team repeats, then sit with five people while they do it and write down the route each one takes. That list is your spread. Set three finish lines on that task, brief everyone from the same page, and give your strongest user the job of building the version the team keeps.
AI Smart Ventures offers AI Training for growing businesses whose teams are learning at very different speeds. Schedule a consultation to map your team’s spread and design a session that holds all three groups.
People Also Read
- AI for Independent Insurance Agencies in 2026
- v0 vs Bolt vs Lovable for Founder-Built Internal Tools in 2026
About the Author
Nicole A. Donnelly is the Founder of AI Smart Ventures and an AI Adoption Specialist with 20 years of experience as a founder and CEO and over a decade leading AI adoption initiatives. She helps businesses integrate artificial intelligence with clarity and confidence, driving innovation and sustainable growth. Nicole has trained over 20,217 professionals in Applied AI, delivered 624 workshops, and worked with close to 1,000 organizations across diverse industries.
Expertise: AI Transformation, AI Strategy, AI Implementation, AI Adoption, Applied AI, Marketing, Business Operations
Disclaimer: This content is for informational purposes only and does not constitute professional business or technology advice. Results vary based on industry, existing systems and implementation commitment. Contact AI Smart Ventures for a consultation regarding your specific situation.
