Introduction
When we discuss artificial intelligence and employment, the obvious question is whether AI will replace jobs. It’s an important question, particularly for graduates entering a labour market in which some of the work traditionally allocated to junior employees can now be done quickly and cheaply by AI. But there is another question that may prove just as important:
If AI does the junior work, how does anyone become senior?
Most careers have traditionally relied upon something resembling a ladder. People enter near the bottom, undertake relatively straightforward work, gain experience, assume greater responsibility and eventually move into positions requiring substantial judgement. The work at the bottom of that ladder is not always exciting. It can be repetitive, administrative and sometimes rather mundane. Yet it serves two purposes: it gets something done, and it teaches somebody how to do the more difficult work that comes later. That second function is easily overlooked.
The work may be routine, but the learning is not
Consider the traditional graduate employee. A junior accountant may begin by reconciling accounts and preparing basic reports. A trainee lawyer might review documents and conduct relatively straightforward research. A graduate analyst may spend hours cleaning data, constructing spreadsheets and producing first drafts of presentations. An experienced professional can look at many of these activities and reasonably conclude that AI could perform much of them, and they may well be right.
The question, however, is not simply whether the task can be automated. We also need to ask what somebody used to learn while performing it. A person who has reconciled hundreds of accounts may begin to recognise when something does not look right. Someone who has reviewed large numbers of legal documents develops instincts about which details matter. An analyst who has spent years working with data gradually develops an understanding of what should be questioned rather than simply accepted. We usually call the result experience, and experience has traditionally been acquired by doing things repeatedly, often under supervision and with relatively limited consequences when mistakes are made.
Higher education has its own career ladders
Universities are not exempt from this problem. Consider the development of an academic career. Doctoral researchers and early-career academics undertake literature reviews, analyse data, prepare teaching materials, mark assessments, respond to students, assist with research projects and perform many other relatively routine academic activities. AI can already help with many of these tasks, and its capabilities will almost certainly increase.
That offers considerable opportunities. Few academics will mourn unnecessary administrative work, and there is little merit in requiring somebody to spend three hours completing a task that can be done in fifteen minutes by AI. But there is an important difference between eliminating unnecessary effort and eliminating developmental experience. Someone who has read hundreds of papers develops a different understanding of a field from someone who has only read AI-generated summaries of them. An academic who has marked hundreds of student assignments learns something about the misconceptions students repeatedly bring to a subject. A researcher who has struggled with messy datasets develops an appreciation of data quality that is difficult to acquire from a perfectly presented AI-generated analysis.
The task may sometimes be inefficient, but the experience gained from doing it may still be valuable. That distinction matters because the case for automation often focuses on the immediate productivity gain while giving much less attention to the longer-term developmental opportunities.
Professional services may face the issue even sooner
The effect may be even more obvious in university professional services. Admissions, finance, human resources, registry, marketing, student support and administration all contain activities that lend themselves to automation. Routine enquiries can be answered automatically, documents can be drafted, applications processed, reports produced and basic analysis undertaken with increasingly little human intervention.
From an institutional perspective, there are obvious advantages. If technology enables a university to deliver the same service with fewer people, particularly when finances are under pressure, leaders cannot ignore that opportunity. Yet universities also need future registrars, finance directors, HR directors, heads of admissions, marketing leaders and chief operating officers. Traditionally, those people developed through increasingly responsible roles within their profession. If some of the lower rungs disappear, the route into those senior positions may become much less obvious.
The danger of a two-rung career ladder
One possibility is that entry-level jobs will not disappear completely but will change significantly. Rather than employing ten graduates to perform relatively routine work, an organisation might employ three graduates who use AI to perform much larger volumes of work. Those three employees may become very productive, but they may also be asked to exercise judgement much earlier in their careers.
AI can generate the first analysis, draft the report and identify the apparent anomalies. The junior employee may then be expected to decide whether the analysis is sensible. That sounds efficient until we ask where the employee acquired the experience needed to make that judgement. We risk creating a rather strange career ladder in which people are expected to move rapidly from beginner to experienced professional, with fewer opportunities to learn gradually in between. The problem is not that people are incapable of developing quickly. It is that judgement usually develops through repeated exposure to real situations, including mistakes, ambiguity and uncertainty. There are few shortcuts to acquiring that experience.
Universities face the problem twice
For universities, there is an additional complication because we face this challenge both as employers and as educators. As employers, universities need to think about how their own academics, administrators and leaders will develop if AI takes over much of the work traditionally performed by a human.
For decades, universities have been told that graduates should leave with knowledge, transferable skills and the ability to learn in the workplace. But what if the workplace itself no longer provides the same opportunities to learn? Graduate employability cannot simply mean teaching students to use AI tools. Students may need much more deliberate exposure to judgement, ambiguity, decision-making and responsibility before they leave university, because the workplace may no longer provide the years required in order to develop these skills.
We should not preserve pointless work
There is an obvious danger in this argument. We could easily conclude that junior employees should continue performing inefficient tasks because previous generations learned by doing them. That would be a mistake. There is no educational virtue in spending an afternoon formatting a spreadsheet, manually transferring information between systems or producing routine documents simply because people used to do so.
If AI can remove tedious work, we should welcome that. The challenge is to distinguish between the work we are happy to eliminate and the learning that we must somehow preserve. That distinction is going to become increasingly important because, in many cases, the task and the developmental experience have historically been bundled together. Once AI separates them, organisations will need to think much more carefully about how the learning is recreated and how their future leaders are developed.
Career development may need to become more deliberate
For much of modern organisational life, career development has partly happened by accident. People did the work in front of them. Over time they saw more cases, encountered more problems, made mistakes, watched experienced colleagues and gradually assumed greater responsibility. The job itself provided a form of apprenticeship, even when nobody described it in those terms.
If AI removes substantial parts of that early-career work, organisations may need to recreate the learning more deliberately. That could mean structured rotations, simulations, shadowing, supervised decision-making, mentoring and giving junior employees responsibility for increasingly difficult cases while experienced colleagues remain closely involved. It may also mean allowing people to do things that AI could technically do more efficiently because there is educational value in a human doing it, at least occasionally.
That may feel uncomfortable in organisations focused heavily on productivity. But optimising every individual task does not necessarily optimise the organisation over the longer term. A system that eliminates almost all novice work may become highly efficient in the short term while weakening its ability to develop the people it will depend on in the future.
The productivity paradox
There is a potentially important paradox here. An organisation may make entirely rational decisions about individual tasks. AI can do this more quickly. AI can do this more cheaply. AI can do this without another member of staff. Each decision may make perfect sense when viewed in isolation.
But repeat those decisions often enough and, ten years later, the organisation may discover that it has removed many of the experiences through which its senior people used to develop. The organisation has become more efficient at producing today’s work while becoming less effective at producing tomorrow’s expertise. That is a very different kind of AI risk from the one that dominates most discussions about automation and employment.
Preserve the learning, not necessarily the job
The question for leaders is therefore not whether we should protect every existing entry-level role. We should not. Jobs have always changed as technology develops, and many routine tasks should disappear. The more important question is what happens to the learning that those tasks once provided.
When universities automate an administrative process, we should ask how the people entering that profession will now acquire the knowledge that used to come from operating it. When academics use AI to remove routine research tasks, we should ask how early-career researchers will develop the instincts that those tasks helped to create. And when we redesign programmes for students entering an AI-enabled labour market, we should think about whether universities themselves need to provide more of the experiences that employers once supplied.
Perhaps the biggest challenge is not that AI will remove the bottom rung of the career ladder. It is that we remove it without recognising that that rung existed and served a very important purpose.
I publish two free newsletters each week
Governance, Leadership & Research
A weekly brief for higher education leaders and researchers examining governance, leadership and research in practice.
https://buff.ly/GhsZqVl
Publishing with Integrity
Examining scholarly publishing through an ethical lens, challenging assumptions and supporting academic career development.
https://buff.ly/WiG1yFb
If you subscribe, you will be alerted whenever I publish a new edition.
About the author
Professor Graham Kendall is Vice-Chancellor of GlobalNxt University in Malaysia and an Emeritus Professor at the University of Nottingham. He has held senior university leadership roles spanning research, academic strategy and institutional management, and has published more than 300 peer-reviewed papers. He writes regularly about higher education leadership, research, artificial intelligence and scholarly publishing.