Degrees Under Pressure: Graduate Employability and Labour-Market Transformation in the Age of AI

1. Introduction

For several decades, higher education policy was sustained by a relatively stable public promise: that university study would improve life chances, strengthen professional identity, and expand access to better work. That promise was never evenly distributed, but it was institutionally powerful. Artificial intelligence has not created the fragility of that promise, yet it has made its instability impossible to ignore. Employers are reorganising tasks, professional boundaries are becoming more fluid, and many forms of entry-level knowledge work are being redefined around software-assisted productivity. Under these conditions, the relationship between degree attainment and employability is becoming more contingent, more unequal, and more politically charged.

The scale of transition is not speculative. The World Economic Forum’s Future of Jobs Report 2025 drew on the perspectives of more than 1,000 global employers representing over 14 million workers across 55 economies and positioned technological change, including AI, among the major forces expected to reshape labour markets through 2030 (World Economic Forum, 2025). The International Labour Organization’s refined 2025 index of occupational exposure to generative AI similarly concluded that about one quarter of global employment falls within exposure gradients linked to potential AI-driven transformation, while also emphasising that few occupations are fully automatable and that job transformation is more likely than total replacement (Gmyrek et al., 2025a; ILO, 2025). These findings matter because they shift the debate away from simplistic replacement narratives and toward the more difficult question of how work is being recomposed.

At the same time, AI use is rapidly becoming normal among students, workers, and firms. OECD data released in January 2026 showed that more than one-third of individuals across OECD countries used generative AI tools in 2025, including 41.1% of people in employment and three-quarters of students aged 16 and over, while firm-level adoption had more than doubled over two years (OECD, 2026a). This suggests that graduates are entering a labour market in which AI is no longer a specialist frontier but an emerging condition of ordinary professional life. If so, universities cannot continue to approach employability as though the post-graduation transition is governed by static occupational scripts or by a simple accumulation of generic soft skills.

This paper argues that graduate employability in the age of AI must be understood as a structural question for higher education rather than a narrow individual challenge. The relevant issue is not whether universities should make students “employable” in a crude instrumental sense. It is whether they can equip graduates to enter work environments characterised by task volatility, intensified signalling pressures, uneven access to digital tools, and new demands for judgment, adaptability, and ethical reasoning. The paper proceeds by first reconsidering the concept of employability, then examining how AI is restructuring graduate labour markets, before turning to the implications for universities and for unequal higher education systems.

2. Rethinking Employability Beyond the Skills Checklist

Employability has long been an unstable concept. Dacre Pool and Sewell (2007) helped make it institutionally actionable by presenting a practical model that connected skills, self-efficacy, emotional intelligence, and career development. That framework remains useful because it recognises that employability is not reducible to academic knowledge alone. Yet the model emerged in a policy environment where graduate employability could still be discussed primarily as an accumulation of personal resources. Subsequent scholarship has shown why this is insufficient.

Holmes (2013) challenged the dominant “possession” view of employability, which assumes that graduates become employable simply by acquiring a stock of attributes. He argued that employability also depends on “position” and “process”: the social location of graduates, the ways they are recognised by employers, and the identity work through which they become legible as professionals. James et al. (2013) similarly showed that the graduate skills debate often overstates the transparency of what employers value, while understating the labour-market contexts in which those skills are interpreted. These insights are highly relevant in the AI era because labour markets are becoming more dependent on signals of adaptability, judgment, and technological fluency that may not be visible in traditional credentials alone.

Tomlinson’s (2017) graduate capital model extends this argument by identifying employability as a combination of human, social, cultural, identity, and psychological resources. The significance of this framework lies in its refusal to treat employability as merely a skills deficit problem. Graduates do not enter the labour market as blank economic actors. They enter with different networks, different cultural confidence, different forms of institutional recognition, and different capacities to narrate their value in competitive settings. AI intensifies the importance of these capitals because it changes what is being signalled in recruitment and in early career performance. It is no longer enough to show that one can produce text, conduct standardised analysis, or complete routine professional tasks if machines can now assist with those outputs cheaply and quickly.

Recent conceptual work by Akkermans, Tomlinson and Anderson (2024) is especially useful here. Their model of initial employability development reframes the graduate transition as an interactive and reciprocal process between individuals and employers, shaped by signalling and social exchange. This matters because employability is often discussed as though universities simply prepare graduates and employers simply receive them. In reality, entry into work depends on ongoing exchanges of information, trust, recognition, and opportunity. AI changes those exchanges. It affects what employers think graduates should already know, how productivity is judged, how quickly junior roles are expected to add value, and which aspects of performance remain visible as human contribution.

Employability, then, should not be reduced to a checklist of transferable skills. In the age of AI, it is better understood as the relationship between graduate capability, institutional credibility, and labour-market conditions. This relationship is inherently unequal. Students with stronger digital access, richer work experience, elite institutional affiliation, or better social capital will often appear more “AI-ready” than those navigating weaker infrastructure or less recognised credentials. Universities that approach employability as a neutral skills agenda risk obscuring these structural asymmetries.

3. AI and the Restructuring of Graduate Work

The most important labour-market effect of AI is not the sudden disappearance of all graduate jobs but the recomposition of tasks within them. The ILO’s 2025 analysis is clear on this point: few occupations are currently fully automatable with generative AI, nearly all jobs retain tasks that require human input, and transformation is more probable than wholesale elimination (ILO, 2025). Yet this should not reassure universities too quickly. For graduates, especially those entering knowledge-intensive occupations, task transformation may be more disruptive than outright replacement because it alters the very work through which professional identity and competence are formed.

Entry-level roles have historically functioned as apprenticeship spaces. Junior analysts, trainees, assistants, and associate professionals learn by handling bounded tasks: preparing first drafts, compiling literature, producing standard reports, summarising cases, checking data, or supporting client communication. These tasks are exactly the ones generative systems can increasingly assist with. The World Economic Forum’s 2026 briefing on entry-level workers captures this emerging tension. It reports that optimism and anxiety coexist among early-career workers, that the perception of entry-level work itself is being redefined, and that many younger workers are learning quickly while also doubting the durability of their skills (World Economic Forum, 2026). In other words, AI is not simply adding a tool to early careers; it is altering the developmental logic of those careers.

This creates a paradox for graduates. On one hand, AI competence can improve productivity and strengthen employability signals. On the other hand, if too much routine cognitive work is delegated to systems, graduates may have fewer opportunities to build the deep tacit knowledge that underpins later expertise. A profession cannot be renewed if its newcomers are excluded from the formative tasks through which judgment matures. Universities should therefore resist simplistic narratives that assume automation of junior work is harmless because higher-order tasks remain human. Higher-order work depends on earlier developmental stages, and those stages are precisely where many graduates first learn to distinguish a plausible output from a sound one.

The problem is compounded by the speed of diffusion. OECD evidence shows not only that individual AI use is rising rapidly, but that adoption is uneven by age, education, and sector (OECD, 2026a). This implies that graduates are entering labour markets with unequal exposure to the tools already shaping workplace expectations. Students who have learned how to use AI critically may adapt more confidently; others may be disadvantaged before recruitment even begins. The uneven spread of AI across firms also means that graduates may encounter radically different work environments depending on sector, company size, and geography. Employability is thus increasingly a matter of navigating differentiated technological ecologies rather than entering a single recognisable graduate labour market.

These shifts also challenge conventional distinctions between academic and vocational education. Traditionally, university graduates have been valued for analytical depth, conceptual abstraction, and non-routine cognitive capacity. Yet if AI begins to perform some elements of abstract symbolic work at scale, universities must clarify what remains distinctive about graduate formation. The answer cannot simply be “more digital skills.” As the World Economic Forum’s 2025 report suggests, employers are also placing weight on broader capacities tied to adaptability, interpretation, and innovation within changing work environments (World Economic Forum, 2025). The university’s contribution remains crucial, but it must now be defended in terms more robust than credential possession alone.

4. Graduate Transitions Under Pressure

The transition from university to work has always been mediated by more than formal learning. Work experience, social networks, institutional prestige, and confidence in recruitment settings all shape outcomes. The importance of these mediating factors becomes sharper under AI conditions because employers are less likely to rely on degrees as self-sufficient signals of readiness. They increasingly seek evidence that graduates can work with digital systems without being led by them, communicate across hybrid teams, evaluate automated outputs, and continue learning as tasks evolve.

Here the distinction between employability and employment becomes important. A graduate may possess substantial capability while still facing weak employment prospects because of sectoral stagnation, economic slowdown, geographic mismatch, or lack of opportunity. This point has been repeatedly obscured in policy discourse. Holmes (2013) warned against conflating graduate attributes with labour-market outcomes, while James et al. (2013) showed that employer demand is often more contingent and ambiguous than employability rhetoric implies. AI makes this distinction even more necessary. A student may be highly capable, digitally fluent, and intellectually strong, yet still face a crowded labour market in which fewer junior roles exist or where employers expect unrealistically immediate productivity.

At the same time, graduates are being asked to narrate their value more aggressively. Tomlinson’s (2017) identity and social capital dimensions are especially relevant here. In an AI-shaped labour market, graduates must often explain what remains distinctly human in their contribution: judgment, ethical discernment, contextual reading, client trust, interdisciplinary interpretation, and the ability to work under uncertainty. These are not trivial additions to employability; they are becoming central to how graduates differentiate themselves in environments where routine cognitive performance is increasingly commodified.

The risk, however, is that such differentiation becomes socially selective. Students from highly resourced institutions are more likely to encounter career coaching, internships, alumni networks, and employer-facing opportunities that help them translate their graduate capital into labour-market visibility. Those from under-resourced institutions may leave with formal qualifications but weaker opportunities to convert them into recognised employability. Bamwesiga’s (2013) study of employers’ understandings of employability in Rwanda is instructive because it underscores that employability is judged not only in terms of technical knowledge but also communication, adaptability, and workplace orientation. In contexts where many graduates must also navigate unemployment, underemployment, or informal work, the demand for these broader capabilities can intensify without corresponding institutional support.

AI may also raise the threshold for “entry level” itself. If employers assume that AI tools have reduced the need for basic junior labour, they may hire fewer early-career staff while expecting those they do hire to operate at a higher level from the outset. This produces a squeezing effect: the first rung of the professional ladder becomes harder to access precisely when young graduates most need transitional roles to build experience. Universities therefore need to recognise that employability support is no longer an optional service added at the edge of the student experience. It is becoming a central institutional responsibility connected to curriculum design, work-based learning, and employer engagement.

5. What Universities Must Now Do

If employability is being restructured rather than simply eroded, universities need responses that are both intellectually serious and institutionally practical. The first requirement is curricular honesty. Courses should identify which tasks in a field are now routinely AI-assisted and which forms of judgment, interpretation, and responsibility remain central to professional practice. This does not mean designing programmes around software training alone. It means helping students understand how work in their discipline is changing, where new risks of deskilling may emerge, and how human value is being redefined.

The second requirement is to embed employability development more deeply within academic programmes. Dacre Pool and Sewell (2007) were right to connect employability with confidence, reflection, and career development rather than isolated skill acquisition. In the current moment, that insight should be expanded. Students need structured opportunities to practise presenting their capabilities, working on real problems, reflecting on technology use, and articulating how disciplinary learning translates into changing workplaces. Work-integrated learning, live projects, supervised internships, clinical simulations, and research-led applied work all remain vital, not because they simply make graduates more marketable, but because they create spaces where knowledge can be tested against the ambiguities of practice.

The third requirement is to treat AI literacy as a component of professional formation rather than a freestanding add-on. OECD’s 2026 digital education outlook argues that generative AI can support education only when guided by clear pedagogical principles (OECD, 2026b). A similar logic applies to employability. Students should learn not merely how to use AI tools, but how to assess their reliability, disclose their use appropriately, detect their limits, and understand their implications for accountability in professional settings. Graduates who can critically evaluate machine-generated material are likely to be more valuable than those who simply know how to prompt efficiently.

Fourth, universities should strengthen career development as a pedagogical and strategic function. Too often, career services operate separately from academic departments, while employability remains rhetorically everyone’s responsibility and practically no one’s priority. Under AI conditions, that separation is increasingly untenable. Departments need labour-market intelligence, employers need more realistic engagement with academic learning, and students need help converting academic achievement into credible career narratives. Graduate outcome data, alumni pathways, and employer feedback should inform curricular review without allowing employer demand alone to dictate educational purpose.

Finally, universities must avoid shrinking themselves into training providers for the AI economy. Employability matters, but a university that abandons critique, civic education, ethical reflection, and disciplinary depth will ultimately weaken the very capacities that make graduates resilient in volatile labour markets. The strongest employability strategy is not a narrow one. It is an educational one: graduates who can interpret complexity, act with integrity, communicate well, collaborate across difference, and continue learning in unstable professional environments are likely to remain valuable even as tasks change.

6. Unequal Contexts, African Realities, and the Politics of Transition

The global future-of-work debate often assumes far more institutional capacity than many universities actually possess. The ILO’s 2025 findings make this clear by showing that exposure to generative AI varies sharply across income groups and that low-income countries face lower exposure in part because of infrastructure constraints, limited digital skills, and the cost of adoption (ILO, 2025). These constraints do not place such systems outside AI change; rather, they shape how that change is experienced. In many African settings, graduates are entering labour markets already characterised by informality, scarce professional openings, and uneven digital infrastructure. AI may create new opportunity niches, but it may also deepen stratification between institutions and graduates with access to tools, networks, and employer confidence and those without them.

This is why employability policy in such contexts must be socially grounded. Bamwesiga’s (2013) Rwandan study reminds us that employers value practical orientation, communication, and workplace adaptability alongside academic learning. These expectations are unlikely to disappear in the AI era; they are likely to intensify. Yet it would be a mistake to respond by calling for universities simply to become more market-driven. Where labour markets themselves are unstable, the university’s responsibility includes helping graduates navigate uncertainty, entrepreneurship, public service, and hybrid career paths, not only formal corporate employment.

There is also a sovereignty question. If AI tools, labour-market platforms, and employability metrics are imported wholesale from dominant digital economies, universities in Africa risk becoming dependent on external standards of value that do not reflect local development priorities. A more defensible strategy would connect graduate employability to national and regional goals: public administration, health systems, education, agriculture, sustainable enterprise, local innovation, and digital public infrastructure. In that frame, employability is not only about insertion into existing markets; it is also about preparing graduates to shape and improve them.

Policy support matters here. Universities alone cannot solve youth unemployment, weak industrial absorption, or technological inequality. But they can become more intentional about graduate transition. This includes stronger employer partnerships, investment in career learning, better graduate tracking, digital inclusion strategies, and serious staff development on AI and work. Without such support, the rhetoric of employability may simply shift more risk onto students while institutions continue to promise outcomes they cannot structurally secure.

7. Discussion

The central mistake in current debate is to imagine that AI has created an employability crisis that can be solved through faster skills updating. The problem is deeper. AI has exposed how fragile the post-university transition already was, how dependent it has always been on social recognition and opportunity structures, and how thin many institutional approaches to employability have become. A narrow “job-ready graduate” agenda will not resolve these pressures. Indeed, it may intensify them by reducing education to immediate labour-market responsiveness just as labour markets themselves are becoming less stable.

A more serious response begins by recognising that graduate employability is both educational and political. It is educational because universities shape the knowledge, judgment, and confidence with which students approach work. It is political because labour markets distribute opportunity unequally, and because technological transitions create winners and losers long before any individual graduate arrives at an interview. The task, therefore, is not to promise certainty, but to design higher education in ways that increase graduates’ capacity to navigate change without abandoning the broader public purposes of the university.

Under AI conditions, the most defensible university position is one that combines realism with intellectual ambition. Realism means acknowledging that a degree alone is no longer a sufficient employment signal and that many traditional graduate pathways are under pressure. Intellectual ambition means refusing to let employability become a euphemism for narrow training. Universities should aim to produce graduates who can enter changing labour markets critically, ethically, and adaptively, not merely those who can reproduce the competencies most easily measured by employers at a single moment in time.

8. Conclusion

Graduate employability in the age of AI cannot be understood as a simple matter of acquiring more technical skills. Artificial intelligence is changing the organisation of work, the structure of entry-level opportunities, the meaning of productivity, and the signals by which graduates are judged. These changes intensify pre-existing inequalities while also creating new demands for adaptability, reflection, and professional judgment. The degree remains important, but it no longer speaks with the same automatic authority it once seemed to carry.

For universities, the response should not be panic and should not be submission. The challenge is to reconnect employability with serious education. That means designing curricula that illuminate changing work, supporting graduates in the difficult transition to first destinations, building credible partnerships without abandoning academic purpose, and recognising that employability is shaped as much by structures of opportunity as by individual attributes. In the end, the strongest graduate in an AI-shaped labour market will not be the one most fluent in fashionable tools, but the one able to combine technological awareness with disciplined thought, social intelligence, ethical responsibility, and a capacity to learn beyond the shelf life of any single platform.

References