We are becoming increasingly efficient at doing the work, but potentially less effective at developing the people who will do it in the future.

Artificial intelligence is changing the way organisations work. But there is a growing problem that deserves much more attention: what happens when AI takes over the tasks through which people traditionally gained their first experience of working?

For generations, starting a career meant doing the less glamorous work. Graduates researched, prepared presentations, drafted reports, analysed information and supported experienced colleagues. They made mistakes, received feedback, developed confidence and gradually took on greater responsibility.

These were not simply jobs to be completed. They were the first rungs on a career ladder.

AI can now produce first drafts of many of these outputs in seconds. Employers have an obvious incentive to use it. Why recruit several junior employees to undertake work that technology can perform faster and potentially more cheaply?

But there is a consequence that could prove considerably more expensive in the long term.

If organisations automate the work through which people gain experience, they risk automating away the development of their future workforce.

The entry-level job is changing

The pressure is already visible in recruitment.

The Institute of Student Employers (ISE) reported that, among its surveyed employers, graduate recruitment fell by 8% in 2024/25, while overall entry-level recruitment declined by 5%. Those employers received an average of 140 applications per graduate vacancy.

Its 2026 recruitment report subsequently found a 4% rise in graduate hiring in 2025/26, although the increase was concentrated among a few large employers. Competition remained intense, at 141 applications per graduate vacancy. The picture is not simply one of disappearing opportunities: early-career roles are changing, with greater emphasis on critical thinking, professional judgement, problem-solving and communication.

There are also warning signs in specialist professions. UK Government cyber security skills research found that, among core cyber job advertisements specifying minimum experience requirements, 16% sought candidates with no more than one year's experience in 2025, down from 25% in 2022.

Recruitment figures alone cannot establish AI as the cause. Indeed, ISE research found that 82% of employers expected no or minimal replacement of early-career roles by AI. Yet an August 2026 Stanford study found that employment among US workers aged 22–25 in AI-exposed occupations stood 19% below where it would have been had it kept pace with less-exposed occupations. The authors stressed that the finding was descriptive, not proof of causation.

The contradiction is becoming apparent: employers need skilled professionals but the opportunities for inexperienced people to become skilled professionals may be diminishing.

You need experience to get experience

Consider a marketing department. A junior employee might previously have researched competitors, prepared campaign drafts, analysed website performance and produced reports. An experienced marketer would review the work, explain what was wrong and gradually entrust the newcomer with more complex assignments.

Now AI can undertake much of that production.

The same applies across professions. AI can assist accountants with financial summaries, lawyers with document research, HR teams with routine queries and developers with code generation.

But completing a task and understanding the task are not the same thing.

A junior accountant needs to learn how to interpret figures, not just produce them. A young lawyer needs to develop judgement, not simply summarise documents. A developer needs to understand systems and diagnose problems, not merely generate code.

AI can produce an answer without necessarily developing the person who needs to understand why that answer is right, wrong or appropriate.

This is the experience paradox. Employers increasingly want people who can contribute immediately, but the traditional opportunities to develop that capability may be shrinking.

The apprenticeship we didn't realise we were losing

Apprenticeship should not be understood simply as a formal training programme or qualification. Much professional development happens informally through observation, participation, feedback and experience.

People learn by doing the work, watching others, making mistakes and gradually developing the confidence to make decisions themselves.

A junior project manager might begin by organising meetings and updating schedules. Over time, they learn to manage competing priorities, handle difficult stakeholders and recognise when a project is going off track.

These experiences develop professional judgement that cannot simply be downloaded into an AI system.

The danger is that organisations retain experienced employees, equip them with AI and reduce the number of people coming through behind them.

For a while, that may look like excellent workforce efficiency. But what happens when those experienced employees leave, retire or move on?

Who replaces them?

An organisation can automate a task. It cannot afford to automate its own future talent pipeline.

We need to rethink the journey from education to employment

Perhaps the bigger issue is that our education and employment systems are still based on a model of working that AI is rapidly changing.

Traditionally, education provided the foundations, employers provided practical experience and people developed their careers through a combination of training and doing the job.

But what happens when the workplace no longer provides the same opportunities to learn?

Schools, colleges and universities can provide knowledge, qualifications and foundational skills. They cannot fully replicate the experience of working in a real organisation, dealing with customers, managing competing priorities, collaborating with colleagues and making decisions where the consequences matter.

We may therefore need to rethink how we educate, train and introduce people to employment.

Teaching young people how to use AI is important, but insufficient. They also need to learn how to question it, challenge its output, apply knowledge, exercise judgement and take responsibility.

There may be a greater role for employer-supported learning, work placements, project-based education, apprenticeships and structured transitions between education and employment.

Perhaps apprenticeships themselves need redesigning. Rather than learning primarily through routine tasks, future apprentices could learn by working alongside AI, evaluating its output, solving problems and progressively taking responsibility for more complex assignments.

This requires closer collaboration between educators and employers. Education cannot be expected to produce fully experienced employees, and employers cannot expect to recruit people who already possess experience they have never been given the opportunity to acquire.

We need a better bridge between learning and working, not simply a different curriculum at either end.

Don't confuse doing less work with developing more capability

Not every repetitive or administrative task is a valuable learning experience. We should not preserve inefficient work simply to keep inexperienced employees occupied.

The opportunity is to distinguish between tasks that can be automated and experiences that must be retained.

If AI prepares a financial report, a junior employee could investigate anomalies, challenge assumptions and explain the commercial implications.

If AI produces a marketing campaign, a newcomer could evaluate its relevance, test its effectiveness and make recommendations.

If AI generates software code, a junior developer could examine its quality, test its security and understand its implications.

The technology handles more of the production, while the person develops the understanding, judgement and responsibility that underpin professional competence.

This suggests a different approach to entry-level employment: less emphasis on performing routine tasks, more emphasis on learning how to evaluate, improve and take responsibility for the work AI produces.

The leadership challenge: Build capability, not just efficiency

Business leaders need to ask a different question when considering AI investment. Not simply, How much work can we automate? but also, What happens to the people who would previously have learned to do that work?

That means redesigning entry-level roles around meaningful AI-enabled work, protecting mentoring and practical learning, creating opportunities to practise decision-making and measuring skills development alongside productivity.

It also requires a change in management thinking. Managers must become developers of capability, not simply allocators of work.

This is an organisational agility issue. Agile organisations continually bring people, processes, technology and work together in the most effective way. AI adoption should therefore be an opportunity to redesign work and develop human capability, not simply a technology-led exercise in reducing costs.

The traditional career ladder may need to change. Progression could increasingly involve different assignments, projects, experiences and responsibilities rather than simply moving through a sequence of job titles.

But the ladder still needs a first rung.

The take-outs: Five questions business leaders should ask

  • Are we automating the work or removing the opportunity to learn? Consider the impact of efficiency decisions on future workforce capability.
  • Where will our next generation of experienced people come from? If entry-level recruitment declines, what replaces traditional routes into professional careers?
  • Are we redesigning junior roles or simply eliminating them? AI should create opportunities for more meaningful work, not just fewer opportunities to participate.
  • Are education and employers working together? New routes into employment must combine foundational knowledge with practical experience and AI capability.
  • Are we measuring tomorrow's capability as well as today's productivity? The return on AI investment should include the organisation's ability to sustain its skills, knowledge and talent pipeline.

Don't pull up the ladder behind you

There is an irony in the AI revolution. Organisations are investing in technology that can make experienced employees more productive, while potentially reducing the opportunities through which inexperienced people become experienced.

The answer is not to preserve yesterday's jobs or resist automation. New roles will emerge, existing ones will change and entirely new capabilities will be needed.

But the transition requires deliberate thought.

We cannot expect the next generation to arrive fully formed, equipped with years of professional judgement, commercial awareness and practical experience. Those qualities have to be developed.

The challenge is not simply to prepare people for an AI-powered workplace. It is to redesign education, learning and work so that people still have a future in it.