What Happens When AI Removes the Work That Teaches Us How to Work?
- 2 days ago
- 12 min read
Written by Rosie Hewat, Founder & CEO of Rosie’s People
Rosie Hewat is a Board and Executive Advisor and the Founder & CEO of Rosie’s People, a leadership and organisational advisory platform. She works with founders, boards, and senior leaders navigating complexity, scale, and high-stakes decision-making across global and regulated environments.
Much of the conversation about AI has focused on productivity. How much time can it save? Which tasks can it automate? How many people can an organisation operate with if AI is doing some of the work? These are important questions, but I believe we may be overlooking a more fundamental one.
What happens when AI removes the very work that teaches people how to become experts?
This question has become increasingly relevant as organisations accelerate AI adoption while simultaneously reducing entry level hiring.
Research cited by the World Economic Forum suggests that entry level positions represented approximately 38.6% of available jobs at the beginning of 2026, down from 44% only a few years earlier. Meanwhile, Stanford Digital Economy Lab research found that early career professionals aged 22 to 25 working in occupations highly exposed to generative AI have experienced a 13% relative decline in employment. Revelio Labs data also suggests that entry level white collar job postings have fallen significantly over the past eighteen months.
The immediate concern is obvious, fewer opportunities for graduates and early career professionals. The longer term concern is less obvious, but potentially more significant.
If fewer people are entering professions at the beginning, where do future experts come from?
Every expert was once a beginner
Every profession has traditionally relied on some form of apprenticeship, whether formal or informal. Lawyers may begin by reviewing contracts, disclosures, and case files. Accountants often start by reconciling accounts, analysing transactions, and understanding financial controls. Recruiters screen hundreds of CVs and conduct countless interviews. Software engineers write code, fix bugs, and investigate why systems fail. Yet simply performing these tasks is not what creates expertise.
A junior lawyer does not automatically learn risk by reading a contract. They learn because a more experienced lawyer reviews their work, challenges their conclusions, highlights what they missed, and explains why it matters. A recruiter develops judgment not merely by conducting interviews, but by learning from hiring managers, senior recruiters, successful placements, failed placements, and the countless decisions that are reviewed and refined over time.
The same principle applies across professions. Experience alone is not enough. What matters is guided experience.
Much of what transforms knowledge into expertise happens outside formal education. Universities, colleges, and training programmes provide important foundations, but they cannot fully replicate the realities of the workplace. Professional capability is developed through exposure, observation, mentorship, coaching, correction, and practical application.
It happens when a junior employee sits alongside someone who has already encountered the problems they are facing. Someone who can explain not only what to do, but why. Someone who can spot mistakes before they become habits. Someone who can provide context, challenge assumptions, and accelerate learning in ways that no textbook, certification, or AI generated answer can fully replicate.
In many professions, the apprenticeship is not simply about learning how to complete a task. It is about learning how to think. It is the gradual transfer of judgment, confidence, context, and experience from one generation of professionals to the next.
In other words, expertise is rarely downloaded. It is developed.
Throughout my career, I have rarely met a senior leader who could not point to a period early in their career when they learned lessons that no textbook could teach. Often, it was the repetitive work, the mistakes, the difficult conversations, and the seemingly mundane tasks that ultimately shaped their judgment. Just as often, however, it was the people they encountered along the way.
Most of us can point to someone who influenced how we think about our profession. Someone who showed us what good looked like. Someone who challenged us, corrected us, encouraged us, or inspired us to aim higher than we otherwise might have done.
More than twenty years ago, one of those people for me was Sylvia Clovey, a Black female HR Director at Community, the trade union where I worked early in my career. It was not simply her technical expertise that left an impression. It was her integrity, her commitment to facts and accuracy, and the way she approached difficult situations with professionalism and fairness. It was also the importance and emphasis she placed on relationships. Watching someone operate at that level helped shape my own understanding of what great HR and People Ops leadership could look like.
These influences matter because careers are not built through knowledge alone. They are built through observation, mentorship, challenge, encouragement, correction, and example. We learn not only from the work itself, but from the people around us. We learn what good judgment looks like. We learn how to navigate complexity. We learn how to handle mistakes. We learn how to lead.
The concern is not that AI removes work.
It is that it may remove some of the experiences, relationships, and opportunities through which professional confidence, judgment, and capability are traditionally built.
After all, AI itself had to be trained by humans in the first place. Those humans learned from other humans before them. Knowledge, judgment, expertise, and professional values have always been transferred from one generation to the next through a combination of experience, mentorship, and practice. Technology may change how that transfer happens, but it does not eliminate the need for it.
The challenge is that many of the activities traditionally associated with early career development are precisely the activities organisations are now seeking to automate.
At first glance, this seems entirely logical. After all, nobody enters a profession because they enjoy repetitive spreadsheets, document reviews, data reconciliation, or administrative work. Yet there is a danger in assuming that work we dislike is therefore work without value.
The grunt work fallacy
One of the most common assumptions surrounding AI is that eliminating routine work automatically creates better opportunities.
The logic appears sound. If a junior employee no longer spends hours gathering information, reviewing documents, preparing reports, or conducting basic research, surely they can spend that time doing more valuable work. Perhaps.
But this argument assumes that repetitive work exists purely to produce an output. Research into expertise development suggests something different. Professor K. Anders Ericsson’s work on deliberate practice demonstrated that mastery emerges through repeated exposure, feedback, correction, and refinement. In cognitive psychology, these repetitions contribute to what is known as schema formation, the mental frameworks that help individuals recognise patterns, identify anomalies, and make increasingly sophisticated decisions.
Expertise is not simply knowledge acquisition. It is the accumulation of thousands of small learning loops.
When a junior accountant manually reviews hundreds of transactions, they are not simply processing data. They are developing pattern recognition.
When a junior lawyer reviews disclosure documents, they are not simply reading files. They are learning how risk presents itself in practice.
When an HR professional handles employee cases, they are not simply following a process. They are developing judgment, empathy, and contextual decision making.
The repetition itself is part of the learning. The work teaches the worker. This creates one of the most important questions facing organisations today, "If AI removes the work, what replaces the learning?"
The disappearing first rung
The evidence suggests this concern is no longer theoretical. Recent labour market research points to a significant contraction in entry level opportunities across many knowledge based professions.
Deloitte’s 2026 enterprise AI research indicates that many organisations are actively reassessing entry level hiring as automation capabilities mature. Stanford’s Digital Economy Lab identified a measurable decline in employment among workers aged 22 to 25 in occupations most exposed to generative AI. Meanwhile, employers across professional services, technology, marketing, finance, and legal sectors continue to seek productivity gains through automation and AI assisted workflows.
Importantly, this shift is not primarily occurring through mass layoffs. Instead, it is happening through slower hiring, smaller graduate intakes, fewer junior positions, and reduced recruitment pipelines.
The first rung of the career ladder is not disappearing overnight, but it does appear to be getting narrower and while this may improve short term efficiency, it raises a difficult strategic question.
If organisations significantly reduce junior hiring today without deliberately redesigning how capability is developed, where will their future managers, specialists, and leaders come from tomorrow?
The hidden recruitment shift
The picture becomes even more complex when we examine how organisations are hiring. Data from UK labour market studies suggests that publicly advertised graduate vacancies have fallen sharply. However, direct employer hiring has not declined at the same rate.
Instead, recruitment is increasingly moving into private channels, graduate programmes, apprenticeship schemes, university partnerships, internal referrals, and direct talent pipelines.
In many ways, the market is becoming less visible rather than disappearing entirely. Part of this shift appears to be a response to the volume of AI assisted applications now flooding public job boards. Employers are increasingly seeking alternative ways to identify talent and reduce noise within the recruitment process.
The challenge is that these routes are often harder to access for individuals who lack professional networks, industry connections, or knowledge of how these systems operate. This creates a new form of inequality, not necessarily an inequality of capability, but an inequality of access.
The experience bottleneck
These shifts are also becoming visible in ways that labour market statistics alone struggle to capture. In my own recruitment work, I have noticed a growing number of candidates applying for roles despite openly acknowledging that they do not meet the requirements. In many cases, they are not expecting to secure the position itself. They are simply searching for an opportunity to gain experience, secure an internship, or find some way of entering a workforce that increasingly demands experience before experience can be obtained.
Increasingly, I am finding that potential is not the problem. Opportunity is. The challenge has become circular. Employers want experience. Graduates need opportunities to gain experience. Yet the pathways that historically bridged that gap appear to be narrowing.
As a result, many young professionals are taking roles completely unrelated to their studies or long term career ambitions simply because those are the opportunities available. For some, this becomes a temporary detour. For others, it becomes a permanent redirection of their careers.
One experience has stayed with me in particular.
A candidate once offered to pay for the opportunity to work within my business. I declined immediately, but the conversation was revealing. It felt less like a negotiation and more like a reflection of how desperate some individuals have become to gain the experience that employers increasingly demand.
When people begin offering to pay for experience rather than being paid to gain it, it may be a sign that the traditional pathways into professional careers are under greater strain than many organisations realise.
The judgment problem
Perhaps the most overlooked challenge is what I would call the judgment problem.
Increasingly, junior professionals are being asked to review, refine, validate, and critique AI generated outputs. On the surface, this appears to represent more advanced work. The reality is more complicated.
Effective judgment requires deep understanding. Understanding often comes from doing. One observation from the research stood out to me more than any other, the editor must first understand the craft of the producer.
In practical terms, it is difficult to evaluate whether something is correct if you have never learned how it is created. Can a junior lawyer confidently challenge an AI generated legal argument if they have never developed the discipline of constructing one themselves? Can a junior accountant identify subtle financial errors if they have never manually worked through the underlying calculations? Can a junior HR professional properly assess an AI generated employee relations recommendation without understanding the realities of human behaviour, organisational politics, and workplace risk?
AI can generate outputs. Human judgment still depends on experience.
The question is whether we are creating enough opportunities for that experience to develop.
This challenge becomes even more complex when viewed through the lens of regulation. If organisations are asking increasingly inexperienced employees to exercise higher levels of judgment, oversight, and accountability, they are doing so at the same time that employment regulation is becoming more demanding and workforce decisions are becoming more scrutinised.
The result is a growing tension between the speed at which technology is changing work and the caution with which organisations are expected to manage people. As AI accelerates change, regulation is increasingly asking organisations to move more deliberately. Navigating both realities simultaneously is becoming one of the defining leadership challenges of the modern workplace.
The regulatory compounding effect
The challenge is further complicated by changes in the regulatory environment.
In the UK, the implementation of the Employment Rights Act 2025 introduces stronger employment protections and additional workforce obligations that many employers expect will increase employment costs.
The result is a tension that many leadership teams are already grappling with. As the cost and complexity of employing people increase, the commercial case for automation becomes stronger. This does not mean organisations are choosing technology over people. But it does mean that workforce decisions are increasingly being influenced by a combination of technological capability, economic pressure, and regulatory change.
Together, these forces may further accelerate the reduction of traditional entry level opportunities.
The leadership pipeline risk
This issue extends far beyond graduates and junior employees. It may ultimately become a leadership challenge. Historically, organisations built future leaders by creating pathways through which people could gain experience, build credibility, and gradually assume greater responsibility.
In our advisory work at Rosie’s People, we increasingly find ourselves helping leadership teams navigate a challenge that sits at the intersection of workforce strategy, technology adoption, and long-term capability development. The question is no longer whether AI should be adopted. Increasingly, the question is how organisations continue to develop capability while work itself is changing.
Because leadership pipelines do not appear automatically. They are built through years of accumulated experience, mentorship, exposure to complexity, and opportunities to practise judgment in progressively more challenging environments.
Research from Gartner and other workforce analysts suggests that many organisations are becoming increasingly concerned about succession planning and leadership development. If organisations reduce junior hiring today, they may not feel the consequences immediately. The impact may emerge three, five, or even ten years later, when there are fewer experienced individuals ready to step into management and leadership positions.
The challenge is not simply replacing tasks. It is sustaining the systems that create capability.
The contrarian view
Of course, there is another side to this argument. Some researchers and business leaders argue that AI is not destroying apprenticeship at all. Instead, they believe it is accelerating it.
From this perspective, AI allows junior employees to bypass years of administrative work and engage with higher value thinking much earlier in their careers. Rather than spending hours preparing data, they spend time interpreting it. Rather than writing first drafts, they evaluate and improve them. Rather than gathering information manually, they focus on decision making.
There is merit to this argument.
Empirical studies from Harvard and Wharton examining what researchers describe as the “skill compression effect” suggest that generative AI can significantly improve the performance of less experienced workers, narrowing the capability gap between novices and experts in certain tasks.
AI can also be an extraordinary learning tool.
It can provide instant feedback, explain complex concepts, simulate real world scenarios, and act as an always available coach. Some organisations are actively redesigning junior roles around these capabilities rather than eliminating them entirely. The question is not whether AI can accelerate learning. The question is whether organisations are intentionally designing for that outcome, or whether they are simply removing work and hoping development takes care of itself.
These are not the same thing.
Towards the AI augmented apprenticeship
Perhaps the answer is not to preserve traditional apprenticeship models exactly as they existed before. Nor is it to eliminate them entirely. Perhaps the future lies somewhere in between.
An AI augmented apprenticeship recognises that routine work still has developmental value while also acknowledging that technology can accelerate learning when used intentionally. Instead of asking how AI can replace junior employees, organisations might ask how AI can help develop them faster. Instead of removing learning loops, they might redesign them. Instead of replacing mentorship, they might strengthen it.
As someone who spends much of my time speaking to employers, candidates, founders, and leadership teams, I increasingly find myself sitting in the middle of a conversation where each group sees a different part of the problem.
Employers are seeking productivity and efficiency. Graduates are seeking opportunity. Governments are investing in skills and apprenticeships. Yet all three groups are reacting to a labour market that is evolving faster than traditional workforce models were designed to accommodate.
The most successful organisations may ultimately be those that understand a simple truth, technology can accelerate capability, but it cannot eliminate the need to build it.
Conclusion
AI is transforming how work gets done, and it is here to stay. What remains less clear is how future expertise will be developed if many of the traditional pathways into professional competence continue to shrink.
Every expert was once a beginner. Every leader was once inexperienced. Every specialist once relied on someone else to teach them how the work really worked.
The question is not whether AI changes how expertise develops, it already has. The question is whether organisations are intentionally designing new pathways to replace the ones they are removing.
Because if we eliminate the first rung of the ladder, we should not be surprised when fewer people reach the top. If fewer people reach the top, the question may not be whether AI changed the workforce, but whether we unintentionally weakened the system that develops the people needed to lead it.
Read more from Rosie Hewat
Rosie Hewat, Founder & CEO of Rosie’s People
Rosie Hewat is a Board and Executive Advisor and Founder & CEO of Rosie’s People, a leadership and organisational advisory platform. A former Group Chief People Officer and Non-Executive Director, she has supported leadership teams and boards operating in high-growth and regulated environments. Rosie is also a trustee and an Executive Contributor to Brainz Magazine, where she writes on leadership, governance, power, and organisational risk.
References:
World Economic Forum (2026), AI-related leadership crisis and workforce development research.
Stanford Digital Economy Lab, The Iceberg Index and workforce exposure to generative AI.
Revelio Labs, entry level labour market and hiring trend analysis.
Deloitte, State of AI in the Enterprise (2026).
Gartner HR Research (2025 to 2026).
CIPD Labour Market Outlook (2026).
UK Employment Rights Act 2025 and implementation guidance.
Ericsson, K. A. et al., The Cambridge Handbook of Expertise and Expert Performance.
PwC Workforce Hopes & Fears Survey.
Institute of Student Employers, ISE Graduate Recruitment Survey.
.jpg)
