Artificial intelligence is changing what people do at work, but there is another question we may need to ask…what did people learn by doing the work that AI is now taking away?

Much of the debate around AI and employment has concentrated understandably on jobs.

Which roles will disappear? Which will change? Which new skills will people need? But recent research points towards a different challenge, one that may have significant implications for organisations managing increasingly multigenerational workforces. AI doesn’t only automate work, it can also automate some of the experiences through which people historically learned how to work.

The work beneath the work

A recent McKinsey article, Building expertise in the age of AI: Who trains the next generation?, highlights an important consequence of changes to entry-level work. Tasks such as research, documentation, data preparation, basic coding and preliminary analysis are increasingly capable of being streamlined or undertaken using AI.

Those tasks can look routine, but McKinsey makes an important observation, they have also traditionally provided the environment in which less-experienced employees develop instincts, practise judgement and gradually acquire expertise. As those activities disappear or change, organisations cannot assume that the old apprenticeship process will continue automatically.

That distinction deserves attention, because sometimes the value of a task isn’t limited to its output, the process of doing it teaches us something.

Efficiency has always changed how we learn

Many of us can probably identify examples from our own working lives. For example, 40 years ago, working in retail banking, computers and email had yet to become part of everyday working life, and as a cashier, one of the basic disciplines was balancing a handwritten cash ledger.

Technology subsequently made that process dramatically faster and more efficient. Few of us would advocate returning to handwritten ledgers, but manually reconciling those numbers also taught something about the mechanics underneath the system, where figures came from, how transactions connected and how to investigate something when the balance was wrong. The technology didn’t simply perform the task faster, it changed the knowledge required to perform it…AI may represent a considerably more powerful version of the same transition.

Three kinds of knowledge?

Perhaps this gives us a useful way of thinking about organisational knowledge. There is technical knowledge, understanding the systems and tools currently used to perform a role. There is underlying knowledge, understanding the mechanics, principles and processes underneath those systems. And there is accumulated judgement, the experience developed over time that helps someone recognise when something doesn’t look right, when the normal process may not apply, or when a different decision is required.

AI has the potential to enhance all three, but organisations may need to think carefully about how all three continue to develop. This becomes particularly important if employees can produce sophisticated outputs increasingly early in their careers without necessarily having experienced all of the steps that once led to them.

Experience and AI may be complements, not competitors

Research from the OECD adds another interesting dimension. Its 2026 paper examining AI in ageing societies found that exposure to AI automation tends to be greater among younger workers and falls with age, with the researchers suggesting that experience tends to complement AI. Importantly, the age pattern becomes less pronounced once education, occupation and country are taken into account.

That qualification matters. It cautions against turning the finding into another generational stereotype, younger people are not inherently “the AI generation”, just as older people are not inherently repositories of wisdom. What matters may be the interaction between technology, experience, role, skills and opportunity.

That also raises one of the questions the ADF is increasingly interested in…When we observe differences between employees of different ages, how much are we really observing age, and how much are we observing experience or tenure?

The human skills may become more valuable, not less

There is another apparent paradox. As AI becomes more capable, some distinctly human capabilities may become more important. The ILO’s recent work on the changing skills landscape identifies growing demand for higher-order cognitive and socioemotional skills alongside digital and AI capabilities. It argues that AI literacy is becoming a foundational capability, but also emphasises adaptability, resilience and human agency.

McKinsey similarly argues that as AI takes on more routine activity, judgement and critical thinking can attract a greater premium. That suggests the workforce challenge isn’t simply, How do we teach people to use AI? it is also, How do people develop the judgement required to use AI well? And that is a very different learning question.

A multigenerational opportunity

This is where the conversation becomes particularly relevant to age-inclusive organisations. We should resist constructing another simplistic exchange in which younger employees provide “digital skills” while older employees provide “experience”…real workplaces are far more complicated than that.

Instead, organisations could ask how expertise moves in every direction. Someone with decades of occupational experience may possess judgement that isn’t documented anywhere. A newer employee may challenge an established process because they haven’t become accustomed to the assumptions behind it. Someone with extensive organisational tenure may understand why a particular system evolved. Someone who joined six months ago may bring substantial expertise from somewhere else. And AI itself may allow all of them to access, interrogate and combine knowledge differently.

That is not simply knowledge transfer…it is knowledge exchange.

Designing the learning that automation removes

McKinsey’s response to the challenge is instructive. It argues that organisations need to connect knowledge management, role design, learning within work and managerial coaching rather than assuming expertise will develop through the same mechanisms it did previously.

That feels like an important principle beyond entry-level work, every time an organisation automates a task, perhaps it should ask two questions:

  • What work have we removed?

  • What learning might we have removed with it?

The answer shouldn’t be to preserve inefficient work simply because previous generations learned through it. It should be to design a better route to the knowledge that work once created.

That might mean simulation, coaching, mentoring, deliberate exposure to complex cases, collaborative problem-solving, knowledge capture or opportunities to work alongside colleagues with different forms of experience. In other words, learning needs to be designed into the new workflow rather than left behind in the old one.

The workforce question behind the technology question

AI will undoubtedly continue to change work, but organisations that focus only on adoption, productivity and technical capability may miss part of the transformation taking place.

The real organisational asset isn’t simply the ability to complete today’s task efficiently, it is the ability to develop the people capable of making tomorrow’s decisions well. That requires technology, it requires learning, and it requires accumulated experience. And increasingly, it may require deliberate connection between people whose knowledge has been developed at different points, in different organisations and through very different working lives.

When we automate a task, we should ask not only what work we have removed, but what learning we may have removed with it, because the future of work will depend not only on what AI knows, iIt will depend on how people continue to learn what they need to know too.