About forty minutes into a keynote a few months ago, a woman near the back put her hand up and asked me the question I get in almost every room.
“We’ve done the training,” she said. “We bought the seats. Why is nobody using it?”
So I asked her what the training had looked like. She described a learning platform, a library of modules, a completion certificate at the end. All sensible. All well-intentioned. Not cheap. And then she said the thing that told me exactly where it had gone wrong.
“Ninety per cent of the team finished it.”
Ninety per cent finished it, and nobody was using it.
Here is the number I have not been able to stop thinking about since.
Boston Consulting Group (BCG) has been running its AI at Work survey for several years now, and in the 2026 edition 86% of respondents said they expect to need significant upskilling within the next five years. Only 36% said they currently feel adequately trained to use AI effectively.
That gap is wide enough on its own. But it is not the finding that matters.
The finding that matters is that the 36% is unchanged since 2025.
Not up. Not down. Unchanged. Across a year in which AI sat on every board agenda in the country, in which budgets were approved and platforms were rolled out and policies were written, the proportion of people who feel properly equipped to use the technology did not move at all.
That is not a training gap. A training gap closes when you spend money on it. This one did not, which means we are looking at something else.
What we are looking at is a trust gap in the making. Because every month that number sits still, your people are drawing their own conclusion about what it means. And the conclusion is not flattering. It is that the organization intends to carry the technology through this transition, and is rather less clear about whether it intends to carry them.
The wave is real, and your people can see it
None of this is anxiety about a hypothetical. The World Economic Forum‘s Future of Jobs Report 2025 estimates that close to 40% of the skills used at work will change by 2030, and that 59 of every 100 workers will need upskilling or reskilling to keep pace. That is not some distant horizon. That is inside the current strategic planning cycle for most of the organizations I work with.
Australian boards can feel it too. The Australian Institute of Company Directors‘s Director Sentiment Index for the first half of 2026, a Roy Morgan survey of 828 company directors, found concerns about artificial intelligence continuing to climb alongside cyber security and regulatory burden.
So the awareness is there at the top and the awareness is there at the bottom. Which leaves an uncomfortable question in the middle, and it is the question that woman at the back of the room was really asking me: if everybody can see it, why is nothing changing?
Why the training did not work
The standard organizational response to a skills gap is a training program. A platform, a course library, a module with a completion certificate and a dashboard that shows you how many people got through it.
The problem is that AI skills are not classroom skills.
Nobody becomes confident with AI by watching videos about it, any more than anyone learns to drive by watching Formula 1. Confidence comes from using it, on your own work, with permission to be a beginner, and with someone nearby who is one step ahead of you. That is why the completion rate and the usage rate can sit so far apart, and why the completion rate is the more comforting number and the less useful one.
EnterpriseWorks reached the same conclusion in their 2026 Workforce Productivity Report, sponsored by monday.com: the organizations pulling ahead are not the ones with the biggest tool stack or the longest course catalogue. They are the ones building practical skills in the flow of work.
Which makes this a leadership design problem, not an L&D procurement problem. And that distinction is the whole article, because the two get solved in completely different places and by completely different people.
Three moves, all cheaper than the platform
If you lead an organization, here is what I would do before you renew that contract.
Make it role-based. “AI training” in the abstract lands nowhere, because nobody’s job is “in the abstract.” Show the finance team what AI does to reconciliation. Show marketing what it does to campaign analysis. Show the operations team what it does to the report they rebuild by hand every Monday. Relevance is what converts curiosity into use.
Build it in the flow of work. One hour, one real task, one small win, repeated, will outperform any curriculum you can buy. The unit of learning here is not a module. It is a task somebody actually had on their list this week.
Protect the practice time. This is the one that gets skipped, and it is the one that decides the outcome. If every hour AI saves is instantly refilled with another meeting, nobody experiments, nobody gets better, and the capability never compounds. People need slack to build capability, and slack is something only you can authorise.
And if you are reading this as an individual
I want to be honest about something, because after more than 5,000 exit interviews I have heard where this ends.
Waiting for your organization to train you is a career strategy from 1996. It assumes an employer who is planning your development ten years ahead, and that employer has largely left the building. This is not a reason to be cynical. It is a reason to take the wheel.
In my S.P.A.R.K.™ Career Development Framework, this sits under the last step, K, Keep Evolving, and the practical toolkit I use for it is what I call the 3Es.
Enhance. Use AI today on the work already in front of you, and start embarrassingly small. Emails. Research. First drafts. The point of a small win is not the win, it is the confidence, and everything else depends on it.
Evaluate. Audit your own role honestly. Which of your tasks are most exposed to automation, which will be augmented rather than replaced, and which are genuinely, stubbornly human? Map your transferable skills before somebody else maps them for you.
Evolve. Use AI as a coach rather than a calculator, and aim your development at the roles that are growing rather than the ones that are shrinking.
The number is a choice
Somebody decided what that 36% would be this year, and somebody will decide what it is next year. It will not be decided by a procurement committee.
So here is the diagnostic I would run tomorrow, and it takes about ten minutes. Ask three people on your team to show you one thing AI did for their work this week. Not what they think about AI. Not whether they finished the module. One real thing, on real work, this week.
Count how many can.
That number is your actual training result, and it is the only one that has ever mattered.
#FutureOfWork #leadership #AI
Thinking about training your team on how to use AI better? Let’s chat. I help take people from fear to confidence and from curiosity to capability.
