Cover
When AI takes the first step
AI is taking away entry-level jobs and valuable experience along with it. How can universities help replace it?
By Seb Murray
15 July 2026
In brief
- AI is automating entry-level "grunt work," threatening the traditional apprenticeship phase where graduates build vital professional judgment.
- While AI boosts initial productivity, it risks deskilling graduates by stripping away the repetition needed to develop expertise.
- Universities must prioritise human-centric cognitive skills and industry collaboration to rebuild the professional learning curve for future leaders.
For graduates entering the labour market this year, the climb has become markedly steeper. Students are applying earlier, completing more internships and attending more careers events than previous cohorts, yet breaking through less often.
By February of their final year, just 27 percent of UK finalists had secured a graduate job, down from 33 percent three years earlier, according to High Fliers Research, an analytics company focussed on graduate and apprenticeship research. Artificial intelligence is accelerating the squeeze, absorbing many of the routine tasks that once formed the backbone of entry-level work, from crunching numbers in finance to combing through contracts in law.
That has turned AI into the main focus of an increasingly unforgiving hiring market, not just in the UK but in economies across the globe. But it obscures a far bigger shift: many of the tasks disappearing first are the very ones that have traditionally forged professional judgement.
Graduates have long earned that judgement by drafting reports, gathering data and sitting in on client meetings. Much of it was grunt work. But it was also where they developed the instincts which classrooms seldom teach.
That raises a bigger question than whether graduates can secure their first job: if AI is stripping away the apprenticeship on which professional careers are built, what replaces it? And what role should universities play?
“Employers now expect AI competence from every graduate, but no one has yet defined what that competence actually is. That ambiguity is the real challenge for universities,” says Mohammad Khalil, Associate Professor of AI and education at the University of Bergen in Norway.
Reskilling or deskilling?
For generations, professional expertise has oftentimes been built the same way: through repetition. Graduates started with routine work, gradually took on more responsibility and, over time, learnt to make increasingly big calls as eventual managers and leaders.
None of those early tasks looked especially glamorous; nor were they expected to. Each one added another layer of judgement and eventually, graduates stopped asking how to do the work and started knowing how to do it.
AI is beginning to compress that learning curve. Increasingly, the first draft, the financial model, or the market analysis can be produced by a chatbot in seconds. That promises a significant productivity dividend. But it also raises an uncomfortable possibility: if the grunt work disappears, so too might the experience that came with it.
Caitlin Bentley, senior lecturer in AI education at King’s College London, has found in her research that people using AI often struggle to recognise whether their own skills are improving or quietly eroding.
“The more serious issue is what our 'deskilling' research is now showing empirically: people often cannot tell whether they are getting better or worse at their own job,” says Bentley.
In one of her experiments involving ship operators, they were given practice dealing with mistakes made by an AI system. When the same problems emerged later in the test, they performed much better. Yet those who performed worse were just as confident in their own abilities, suggesting AI can make it harder for people to recognise when their own expertise is beginning to disappear.
Bentley says: “Early-career workers in our data reported the highest concern about this, because they are the ones entering the workforce now with the least knowledge and experience to fall back on.” They may appear more productive, and may even feel more capable. But they lack the experience to recognise what they are missing.
Thinking about thinking
The irony is that AI is making human skills more valuable, not less. As chatbots become better at churning out first drafts, value is shifting to the people asking the right questions, spotting the mistakes and exercising judgement.
The QS World Future Skills Index 2027, drawing on responses from more than 92,000 employers across 89 economies, found companies report their biggest skill gaps in human cognitive skills and leadership. Universities are producing graduates with strong technical foundations, but what many employers believe is missing are the capabilities that are far harder to automate: judgement, communication and decision-making.
“The biggest change AI is bringing to employer expectations is the rising premium on distinctly human skills in an era of machine-generated content,” says Xue Zhou, Dean of AI at the University of Leicester in England. “Employers are shifting their focus from what graduates know to how they think, verify and collaborate with intelligent tools.”
She argues that students need domain knowledge to challenge AI and not simply accept its answers as gospel. They also need what she calls “metacognitive skills” – essentially the ability to think about their own thinking – so they can work alongside AI, design better workflows and generate new ideas, rather than simply consume machine-generated ones.
Andrew Crisp, a British higher education consultant, believes employers are now looking well beyond AI literacy. “Human skills remain just as important, and will continue to be so,” he says. That means graduates need to communicate clearly and tell compelling stories with data.
“Employers are also likely to want agile graduates, those that can see there is more than one way to solve a problem, graduates who can live with and embrace uncertainty, and graduates who can imagine.”
Yesterday’s labour market
If the challenge is that graduates are losing the opportunity to develop judgement, then adding another AI module to the university curriculum will not solve it. The model of learning itself will need to change.
At IESE Business School in Barcelona, students begin learning AI before they even arrive on campus. Then, weekly “AI pills” run throughout their degree programmes. But the school has also doubled down on the things technology cannot replace.
IESE has introduced device-free classrooms to force students to think through case studies without relying on laptops or chatbots, alongside a communication bootcamp designed to strengthen the skills recruiters consistently rank among their highest priorities.
“Business schools should not just teach the tools; they should form the judgement and character to use them well,” says Isabel Estalella, Admissions Director at IESE.
The University of Leicester’s Zhou believes assessment also needs to change. Universities should stop judging students solely on the final answer and instead assess how they arrived at it. “The process of learning – making the use of AI transparent, intentional and efficient – is key,” she says.
Zhou also argues that universities need to create far more opportunities for students to tackle real business problems, where AI becomes one tool among many rather than the answer itself.
That means resisting the temptation to let AI become a shortcut around learning.
“Unfortunately, the ‘use genAI without learning the essential skills and core knowledge’ approach can only take our students so far before negatively impacting their ability to master their discipline,” says Sophie Rutschmann, Associate Provost for Digitally Enhanced Learning and Teaching at Imperial College London.
At King’s College, Bentley takes the argument further: universities should resist defining their purpose simply as producing graduates ready for today’s AI tools.
“If universities define their role purely as producing employable, qualified graduates for an AI-driven economy, they will end up teaching for yesterday’s labour market,” she says. “Employability is not the wrong goal, but it’s an insufficient one, and it’s not really the university’s distinctive contribution.”
She argues universities should instead create spaces where students openly debate what AI is doing to work, expertise and society, and argue about what good AI use looks like.
Rebuilding old systems
Teaching students how to use AI will not be enough if they graduate with fewer opportunities to develop professional judgement. Degrees may need to rely less on classroom teaching and far more on practical projects, such as industry placements that provide exposure to real business problems.
That will require universities and employers to work more closely together. If AI is hollowing out the apprenticeship on which young professionals have long depended, employers may have to start rebuilding it, and invest more heavily in training, mentoring and job rotations.
“Universities should not rely only on their own staff to adapt. They need an open discussion with industry to examine which skills are actually needed,” says Professor Khalil at the University of Bergen.
He adds: “Universities should map the competencies expected from graduates, and the goals that follow from them. What was known and taught in the pre-AI era is simply different now. Generative AI, large language models, and now agentic AI have transformed almost everything.”
The QS World Future Skills Index 2027 reaches much the same conclusion. Preparing graduates for an increasingly AI-driven economy will demand closer ties between universities and employers, more learning by doing, and courses that keep pace with the labour market to boot.
“We cannot simply teach subject expertise but have to provide our students with the skills and knowledge they will need to navigate this new world,” says Imperial College’s Rutschmann.
The irony is that AI may help graduates become more productive on day one. But productivity was never solely the point of those early years; experience was too. Unless universities and employers rethink how it is earned, AI could leave organisations with fewer people capable of becoming leaders further down the line.
MEET THE AUTHOR
Seb Murray is a journalist and editor who writes often for the Financial Times and has written for The Times, The Guardian, The Economist, The Evening Standard and BBC Worklife. He focuses on higher education and global business. He also produces a wide range of content for a range of corporate and academic institutions. Seb is also a recognised expert on higher education and speaks at international conferences.