On February 22nd of this year an article named “The 2028 Global Intelligence Crisis” was published by Citrini Research, foretelling the grim and systematic demise of our economies at the hands of Artificial Intelligence1. A focal ingredient in their dystopian creation was a “white collar employment crisis”; a fear that is far from exclusive to the writers of the article, but not far from the realm of possibility.
This depressing story caught fire; amassing hundreds of millions of views across various media platforms and proceeded to burn out a chunk of the stock market. It claimed to be “a scenario, not a prediction” – a disclaimer that, in the frenzy, seemed to be overlooked by many.
Somewhat fortunately – in the context of this article – despite what feats of fiction by AI sycophants or catastrophising fearmongers could lead you to believe, accurately predicting the impact AI will have on the labour market is nigh impossible.
Unfortunately, however, as was clearly demonstrated by the reception to the Citrini article, uncertainty itself is enough to cause tangible economic ramifications. This economic principle, that expectations have huge power over the economy, may also be the key to understanding why the impacts of AI are intuitive, but not yet evident in labour market data.
Isolating the shock: A counterfactual framework
For years, universities have been the toll booths on the fast lane to “white-collar” employment. So, if there is, in the words of Citrini Research, a “human intelligence displacement spiral”, the value of a degree would be at best unclear, and at worst? Well, reduced to an ‘experience’ worth £53,000 of debt for the average student in the UK2.
Thus, as students ourselves, we are uniquely positioned, and incentivised, to answer the question: will generative AI erode the value of a degree in the labour market?
In order to effectively answer this question, however, we must first answer the deceptively simple question that is: what value does a degree actually provide in the labour market?
Though arguments can be made for a great list of attributes, we propose an encompassing value trifecta: signalling, skills, and scarcity. To isolate the impact of generative AI on this three-pronged criteria we assess the value of a degree in a world with AI and compare it to the value a degree would possess in a hypothetical world in which AI does not exist. A comparison that, rather conveniently – although equally disconcertingly – does not require much imagination: only the recollection of reality merely a handful of years ago.
Signalling
Firstly Signalling, this is the idea that a degree, and its classification, may no longer credibly indicate the calibre of its holder. In a world without AI, degrees are a symbol of knowledge, hard work, and differentiation. It is widely accepted that generative AI is making the degree-attainment process easier, be it through research, problem solving, writing, or pedagogical assistance; it is incontestable that to some extent generative AI is helpful – and that is enough.
The likely outcome is grade inflation, or – a more obscure and less solvable issue – grades reflecting not ability, but how much a student pays for their AI subscription. One must note that grade inflation precedes AI ubiquity, but the notion that, in the current state of academia, the issue is not, and will not be, in any way exacerbated by AI is fanciful.
The fear is that, while the academic system does not adapt, an employer may no longer trust a degree, or its classification, to be a reliable signal of a candidate’s true abilities. Students currently applying to internships and jobs are feeling the materialisation of this value erosion first-hand, as more emphasis is placed on other means of assessing capability in the selection process. If a degree’s credibility is lost, is not too its value?
Skills
Secondly, students are facing the macabre reality that AI is causing their own skills atrophy. Following research by MIT, the use of AI is associated with “weaker neural connectivity”. More tangibly, over the course of the study, “LLM users consistently underperformed at neural, linguistic, and behavioral levels”.3
Consider a tool that could make you do your work faster, better, and most crucially, easier than ever previously possible. In the world where this tool exists - the world we find ourselves in today - speed and quality are good, the issue is the ease. AI allows us as humans to circumnavigate the struggle of learning. But this isn’t just any struggle. This is the deliberate and productive struggle that builds durable learning, the very essence of academia.
It is undeniable that an expert in a field utilising AI will be able to leverage its productivity benefits more effectively through superior judgement, and therefore degree holders will be in an advantageous position.
However, it is hard to place where value is created in the case where: a student receives an assignment and puts the brief into an LLM along with the appropriate learning materials; asks AI what topic they should focus on; then asks AI to find them data and relevant papers – which the student naturally asks AI to interpret for them; then uploads the marking criteria and asks AI to produce a draft of a strong assignment which they then copy into their own words.
Is it Ethical? No. Are learning outcomes durable, or even present? Unlikely. But is it effective? Absolutely.
If, as a result of AI, academia no longer produces durable and valuable learning, there is the threat of an ouroboros-like cycle where the creation of the judgement that allows for the leveraging of AI capabilities is hindered by the use of AI itself.
Over time, even if we recognise the long-term costs, natural human myopia could still render AI a crutch, not the enhancing agent it has the potential to be. One could also argue that if productivity is gained purely from Artificial Intelligence and human capital does not improve in tandem, then those productivity gains are precisely that: Artificial.
If the skills of those who hold a degree continue to wither away, then by extension so will the value of the degree itself.
Scarcity
Thirdly, AI has caused a lack of scarcity for the very skills that a degree seeks to foster. Coding, subject matter knowledge, writing, research even reasoning. These skills are fundamental to the identity of today’s scholar and are provided to an increasingly high standard by AI.
Consider a firm’s perspective: before AI, they needed to hire people with degree-level skills to handle data in Excel, create slides, or code. Now they can equip one senior member of their workforce with an LLM and do these jobs for a fraction of the price and possibly at greater efficiency.
This affects graduates twofold: reducing the demand for their skillset simply due to the availability of substitutes; and the second blow is dealt directly to the career ladder, where a diamond shaped employee structure may prevent graduates from performing the entry level work necessary to gain the experience required for career progression.
Perhaps the decline in graduate hiring will reverse when firms realise that if no one hires graduates now, there will be no managers to hire down the line.
However, the job market may fall victim to a classic economic issue referred to as the ‘free-rider’ problem. Graduate hiring in this case could be thought of as a good such as street lighting: everyone wants it, no one wants to pay for it.
Economic theory suggests that such a problem requires regulation; given that a completely free market system struggles to fund lampposts. After all, there’s no guarantee a graduate hire will remain at your firm – rendering return on investment uncertain – and your firm will benefit most if everyone else hires the graduates.
One cannot understate, however, the importance of historical precedent regarding technological progress: consistently and considerably greater job creation than loss in the long term. Although, an economic climate that may be unconducive to expansion, coupled with WEF data estimating that two fifths of the average worker’s skillset may be obsolete by 2030 adds weight to the idea that, in the medium run, a graduate entering the job market may be in for a bumpy ride 4.
What does the data say?
AI sentiment
A recent study by Kings College of London University provides direct insight into the fear surrounding the integration of AI into the labour market5. Interestingly, public sentiment surrounding AI mirrors AI functionality: at face value, the response looks good, but when we examine the finer details, that illusion quickly fades.
When questioned on whether AI is good for “humanity”, University students were recorded to be “more positive than the wider public”. However, wider economic impact seems to be the extent of graduate’s ‘AI-optimism’ – though the King’s research team seems to have missed a student’s liking of AI when deadlines are near and word documents empty.
Fears that the respondent’s own job was going to be replaced by AI were shown to be considerably more widespread among graduate workers than non-graduate workers. Opinions regarding the causes of decreased entry-level job opportunities are also demographically split; with university students being “nearly twice as likely than the wider public” to point out AI as the culprit.
Labour market data
Though the extent as to which AI is responsible is up for debate, the reality of an ever-tightening labour market for young graduates, is not.
Data collected by the job search engine Adzuna reveals the mechanism behind graduate desperation: by July 2025, as was highlighted by the Financial Times (interpreting Adzuna’s data), degree-requiring entry-level job postings in the UK had fallen by approximately 67% from their 2022 levels – notably, the year that ChatGPT was released6; as of February 2026, graduate vacancies are “down -45%” year over year, falling to their lowest level on (Adzuna’s) record7.
These startling figures are contrasted by a -4.4% year on year fall in overall entry-level job postings. This indicates that a degree’s power as a guarantor and stabiliser of job outcomes is faltering, bringing graduate optimism and opportunity down with it.
The data-backed narrative
In the US, amidst similar sentiment and almost identical labour market stress, recent reports by Elise Gould and Joe Fast of the Economic Policy Institute indicate that blaming AI as the sole culprit is a stance that is unsupported by the data 8.
As the labour market impacts seem to be spread across most sectors, the think-tank states that the “culprit is not a structural change in the economy like AI” and instead relate these symptoms to “a labor market in which employers are hiring less and workers are holding on to the jobs they have”.
At first glance, the reasoning seems sound, and evidenced. Considering expectations and the breadth of AI applications, however, the numbers do not seem so opposed to the AI-driven narrative.
AI use-cases seem to be cross cutting through almost all sectors – and its expected use-cases even more so. It follows that AI productivity increases, or the expectation of such increases, could reduce the need, or want for new hires across industries. The caveat being that we should observe a more pronounced impact on non-labour intensive, ‘intelligence’ sectors.
Furthermore, workers holding on tightly to current positions may also be an issue exacerbated by AI: as Gould and Fast highlight, “Quits are higher when workers are confident, they will find better job opportunities”. In which case, it also follows that fear of AI replacement may well be a factor causing workers to hold on to their jobs “more so than any point in the past 10 years.”
Therefore, one must question whether the behaviour of employers and employees are also a symptom, and not the disease.
Labour market shocks: Historical precedent
Though easy to do, we must be wary of blindly buying into age-old clichés of mass replacement and unemployment, or any other catastrophised outcomes of this vein. After all, they have accompanied every technological revolution preceding this one, and have subsequently been spectacularly refuted. Conversely, hoping that AI provides the productivity growth necessary to create more jobs than it displaces does not seem like an apt strategy.
As has been demonstrated by previous labour market shocks, the extent of their impact is largely a question of concentration and labour mobility. If the economy is assumed to have perfect labour mobility, when demand for labour is reduced in one sector, then – like pushing down a corner of a waterbed – obsolete labour will flow to the other sectors of the economy.
This is well illustrated by the movement of typists to other positions requiring transferrable skills, such as that of a secretary, following the introduction of word-processors. The “China-Shock”9, on the other hand, pointed out the flaws in a ‘waterbed’ economic model. The research paper showed that the crater left in the geographically and occupationally immobile US manufacturing sector, a consequence of China’s aggressive export domination, was not easily filled. In fact, it is still yet to be made whole.
In such cases where labour does not move fluidly, economists reason-away concentrated economic pain by pointing to a net-positive impact where the economies become more productive, and people richer (seen in a literal sense in the work of Autor, demonstrating that cheaper imports from China considerably raised the purchasing power of the average American). A concept generally quoted as something like “a rising tide that lifts all ships”.
The thing about a rising tide however, especially one rising at the progress rate of AI and where it looks as if at any moment your employer may throw you overboard, is that you’d better know how to swim.
Adaptation
The question therefore is not, will AI have an impact? But instead, will the impact be concentrated, and the labour transition fluid?
It is unlikely that Chat-GPT will be doing our plumbing any time soon and hiring in labour intensive sectors such as healthcare seems to be holding steady, thus, intuition and data indicate that there will likely be an element of concentration within this impending AI ‘shock’.
This time, however, it is the ‘knowledge’ sector that finds itself on the chopping block, and young workers historically tend to feel the impacts of labour market stress more acutely. Graduates armed only with their own ‘knowledge’ – which is also under threat – and limited experience, seem to already be knee deep in the ‘rising tide’, and first up on the chopping block.
Thus, one must recognise, that the adaptation of academia and the student is the necessary condition to preserve the value of a degree in the labour market.
The onus is on higher education institutions to once again incentivise the development of meaningful human capital – a precarious balancing act considering the simultaneous need to promote AI literacy – and on students to make active choices to preserve their own learning outcomes. The latter of which will almost certainly come at the cost of ease and efficiency, and possibly at the cost of academic performance.
Outsourcing to AI, or meaningful skills development. Ease, efficiency and better grades, or deliberate difficulty, inefficiency and worse academic achievement. The path of least versus most resistance. Every student now faces this dilemma, and the value of their degree rests on their choice.
On the other hand, as AI continues to progress rapidly, bolstered by trillions worth of investment10, a staggering contrast is seen in academia, being an underfunded industry that, for various reasons, has been shown to be slow to adapt to technological change11. For graduates' skills to remain relevant, core academic foundations must be altered or altogether removed.
Universities must adapt testing to once again make a degree, and its classification, reflect ability; questions must be asked about the value of coursework, for example. Universities must also change curriculum formats to incentivise true knowledge formation, not hollow performance. Students and academia may need to become something entirely different to provide skills that will be in demand as AI progresses.
Amidst such uncertainty and speculation, two things are for certain. 1) There is no time for denial, only dynamism: students and academic institutions alike must act promptly if they want to remain competitive. 2) If AI is allowed to erode the value provided by a degree, the already diminishing labour market outcomes of degree holders will continue to erode.
Final Thought
In every technological revolution, there have been winners and losers. This time, a degree alone may not be enough to separate the two.
Footnotes
-
Citrini Research, The 2028 Global Intelligence Crisis (opens in a new tab), February 2026. ↩
-
Paul Bolton, House of Commons Library, Student Loan Statistics (opens in a new tab), December 2025. ↩
-
MIT, Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task (opens in a new tab), December 2025. ↩
-
World Economic Forum, The Future of Jobs Report 2025 (opens in a new tab), 2025. ↩
-
King's College London, One in five Britons think AI will create civil unrest, study finds (opens in a new tab), 2026. ↩
-
Financial Times, Is AI killing graduate jobs? (opens in a new tab), July 2025. ↩
-
Adzuna, Job Market Report (opens in a new tab), 23 February 2026. ↩
-
Economic Policy Institute, Class of 2026: A depressed hires rate is a major cause of labor market weakness for young college graduates (opens in a new tab), May 2026. ↩
-
David H. Autor, David Dorn, and Gordon H. Hanson, Annual Review of Economics, The China Shock: Learning from Labor-Market Adjustment to Large Changes in Trade (opens in a new tab), 2016. ↩
-
Gartner, Gartner Says Worldwide AI Spending Will Total $2.5 Trillion in 2026 (opens in a new tab), January 2026. ↩
-
Springer, Unveiling the barriers to digital transformation in higher education institutions: a systematic literature review (opens in a new tab), 2025. ↩