Here is a worry I hear a lot, usually said quietly and a little reluctantly. “I think I’ve falling behind on AI and I don’t know what to do.” Sometimes it is a senior person who has not quite found the time. Sometimes it is someone watching a colleague who seems to have raced ahead. And often, sitting just underneath it, is a second thought that makes the first one worse: if a much better tool is going to land in a year anyway, is there any point putting in the effort now?
We need to take both of those worries seriously, because they are reasonable. But I also think they are based on a misreading of what is actually going on, and of what it is people are really learning when they learn to work with AI.
The short version is this. The lag is mostly not a personal failing. And the learning you do today is not wasted when the tools change, because most of it is not about the tools at all.
Let’s look at those one at a time.
The lag is mostly about conditions, not character
It is tempting to explain the gap between confident AI users and everyone else as a difference in aptitude. Some people take to new technology, some do not, and that is just how it is. It is a tidy explanation. It is also, on the evidence, mostly wrong, and believing it leads employers to do the wrong things.
Microsoft’s 2026 Work Trend Index, published in May, surveyed 20,000 people who use AI at work across ten countries including the UK. (The featured image for this post comes from their report.) It found something striking. When the researchers tested 29 different factors against the real value people get from AI, organisational factors such as culture, manager support and talent practices accounted for more than twice the impact of individual factors like personal mindset and attitude.
The single strongest factor of all was the organisation’s AI culture, which the report puts at roughly two and a half times stronger a signal than the top individual factor.
So whether a given person gets value out of AI depends far more on the environment their employer has built around them than on how naturally they take to the technology. The “good with AI” colleague is, more often than not, a person who happens to be working in conditions that allow it.
And this is not just one vendor’s finding. It shows up wherever people look. Gallup’s February 2026 survey of more than 23,000 US employees found that while individuals who use AI frequently report real productivity gains, only about one in ten employees in AI-adopting organisations strongly agree that AI has actually changed how work gets done across their organisation. The benefit is stuck at the level of individual tasks because the surrounding systems, the workflows and roles and processes, have not been redesigned.
McKinsey, looking at the same question, has been blunt about where the bottleneck sits: the biggest barrier to scaling AI is not employees, who are ready, but leaders, who are not moving fast enough to change the conditions around them.
So when someone says “I’m falling behind”, the honest response is usually not “you need to try harder”. It is “what has your organisation actually done to make this possible?”
In our own training and advisory work this pattern is very familiar. The people who describe themselves as behind are very rarely incapable or uninterested. They are, almost always, people who have been handed a tool and no time, no examples to follow and no sense that experimenting with it is a legitimate use of their working day. Give the same people a different set of conditions and the so-called aptitude gap tends to shrink dramatically.
Why capable people still hold back
There is a deeper reason the lag persists, and it has nothing to do with willingness. The Microsoft report gives it a name, the Transformation Paradox, and once you see it you cannot unsee it.
Look at three figures from the survey.
- Around 65% of AI users say they fear falling behind if they do not adapt quickly.
- Yet 45% say it feels safer to focus on hitting their current goals than to redesign how they work with AI.
- And only 13% say they are rewarded for reinventing their work with AI when the results do not immediately follow.
Put those together and the behaviour stops looking like reluctance and starts looking like good sense. Learning to use AI well is not a matter of watching a demo. It means experimenting, getting things wrong, and working slower for a while as you rebuild a process you used to be able to do on autopilot. If your organisation only notices and rewards this quarter’s output, the rational move is to keep your head down and deliver the old way. The very same pressure that makes people anxious about AI is also quietly punishing the experimentation that would cure the anxiety.
Other research describes the same trap from a different angle. ManpowerGroup’s 2026 Global Talent Barometer found that regular AI use among workers has climbed, while confidence in using the technology has actually fallen. People are using more AI and feeling less sure of themselves while they do it, which is exactly what you would expect when adoption is running ahead of any real support. A separate UK-focused survey found a clear majority of professionals saying their organisation had given them no adequate guidance on using AI effectively at all.
This is the important shift in thinking. The lag is not a workforce problem to be solved by motivating individuals. It is a systems problem. And systems do not fix themselves. They have to be redesigned, deliberately, by the people who own them.
What employers can actually do
The good news is that the fixes are not exotic. They do not need a new budget line or a transformation programme with a logo. They need leaders and managers to make a few decisions and then back them up. Three stand out:
- The first is to give people genuinely protected time to learn. Not a lunchtime webinar bolted onto an unchanged workload, but real room to experiment without it counting against their delivery. The 45% who play it safe are responding to an honest signal about what their organisation values. If you want different behaviour, you have to change the signal, and time is the clearest signal there is.
- The second is for managers to use the tools visibly themselves. This one has hard evidence behind it. A separate Microsoft study of 1,800 workers found that when managers actively modelled AI use in their own work, their employees reported a 17-point lift in the value they got from AI and a 30-point lift in trust in AI agents. People calibrate what is acceptable and worthwhile by watching their manager, not by reading the all-staff email. A manager who quietly avoids the tools is teaching their team to do the same, whatever the official message says.
- The third is to reward the reinvention, not only the result. If improving a process by experimenting with AI is invisible in how someone is appraised, it will stay a hobby for the few rather than a habit for the many. The most advanced AI users in the Microsoft research were roughly twice as likely as everyone else to say their reinvention of work was recognised regardless of the immediate outcome. That is not a coincidence. It is cause and effect.
None of this is about pushing AI on people who do not want it. It is about removing the entirely rational reasons that capable, willing people currently have for holding back.
Why today’s learning is not wasted
Now to the second worry, the one about the better tool that is surely coming. If today’s AI assistant will look basic in a year, why pour effort into learning it?
Because most of what people are really learning is not the tool.
The Microsoft report makes this point well, and it matches what we see when we train people. As AI takes on more of the execution, the human premium shifts onto judgement:
- Knowing what good output actually looks like.
- Deciding what is sensible to delegate and what you need to keep your hands on.
- Treating an AI answer as a first draft to be interrogated rather than a verdict to be accepted.
In the survey, 86% of AI users said they already do exactly that, treating AI output as a starting point and staying responsible for the thinking.
That judgement does not expire when the software updates. Someone who has spent six months learning to spot a confidently wrong answer, to sense when a task is a poor fit for AI, and to match the right kind of help to the right kind of work, carries every bit of that straight into next year’s more capable tools. They are not starting again. They are starting from a higher floor.
The person who waited, by contrast, meets the better tool with none of that hard-won instinct and a steeper climb ahead of them, because more capable tools fail in subtler, harder-to-catch ways, not more obvious ones.
This is exactly what the image at the top of this post is about. It sets out four modes of working with AI, from delegation through collaboration and asking to exploration, each suited to a different kind of task, each with its own thing to watch out for. The genuinely useful skill is not memorising any one tool. It is learning to ask, before you start, which mode this particular task calls for and how sensitive the information involved is. That habit of matching the mode to the task is the durable skill. It will still be the durable skill when the current tools are a memory.
So the learning curve people are climbing now is not a curve they will have to climb again. It is the groundwork.
The point we teach all our clients
The next generation of AI tools is coming whether organisations are ready for it or not. The ones that end up using it well will not be the ones that happened to hire the most naturally gifted staff. They will be the ones that, starting now, gave their people three things: the time to learn, the visible example of leaders doing it too, and incentives that reward the effort rather than quietly penalising it.
For anyone who feels they are behind, the most useful thing to know is that the feeling is probably telling you something true about your conditions rather than something true about you. And for the leaders and managers reading this, that is rather good news, because conditions are the one part of this you can actually change.
Learning today’s tools was never really about today’s tools. It is the groundwork for everything that comes next.
