Universities UK’s new Future Jobs Roadmap arrives at an awkward but important moment. Employers continue to say that they need graduates, yet many graduates are struggling to secure the first meaningful step into professional work. At the same time, artificial intelligence is beginning to alter the tasks that once formed the entry-level jobs through which graduates learned how to become experienced professionals.

The roadmap is therefore addressing two problems at once. The first is familiar: how can universities, employers and government help graduates develop the skills and experience employers want? The second is much less settled: what will those graduate jobs actually look like when AI systems can perform an increasing range of research, coding, drafting and analytical tasks?

The danger is that the first question will be answered while the second is treated as if it were already known.

What does the roadmap propose?

Future Jobs is one of four projects within Universities UK’s wider Future Universities programme. The roadmap is based on roundtable discussions involving employers, sector bodies and local authorities, followed by a survey of 500 business leaders. It identifies three central issues: a gap in graduate work readiness, the rapid and unpredictable pace of change, and the need to improve employer–university partnerships [1].

Its recommendations include expanding work-based learning, improving students’ access to AI and its workplace applications, providing universal lifelong careers support, developing employer co-designed short learning modules, creating graduate schemes for small and medium-sized enterprises, making it easier for employers to engage with local universities, and reducing the cost of hiring entry-level graduates [1].

The headline ambition is that every undergraduate in England should have access to meaningful work-based learning by 2035, with an interim target of 50 per cent by 2030. The Times Higher Education estimates that this would mean around 1.4 million undergraduates undertaking work experience each year by the middle of the next decade [2]. The roadmap also calls for access to AI tools for every undergraduate and an “AI trailblazer” on every course.

The employer case for expansion is clear. Universities UK reports that 81 per cent of employers support maintaining or increasing undergraduate numbers to meet future skills demand. Yet only 35 per cent say that graduates entering the workforce have all the skills they reasonably expect. Seventy-nine per cent support further work experience [1]. In other words, employers do not generally think that Britain has too many graduates. They think that graduates need more opportunities to develop confidence and practical capability.

That is a useful corrective to the recurring claim that higher education has expanded too far. But it also raises a question that cannot be answered by universities alone: if employers want work-ready graduates, are employers prepared to provide the workplaces in which people become work-ready?

Work experience is necessary but not costless

Work-based learning can take many forms: placements, sandwich years, job shadowing, project work, virtual internships and shorter periods of employer engagement. It can be valuable, particularly for students who have had limited access to professional networks. It may help students understand how knowledge travels into practice and allow employers to see capability that is not visible in a conventional application.

But making work-based learning universal is not simply a matter of adding a requirement to university courses. Placements have to be available, meaningful, supervised and accessible. They must not become a source of cheap or unpaid labour. Universities will also need to recognise that many students already work while studying, often in retail, hospitality or care. Those jobs may not be classified as graduate employment, but they develop timekeeping, communication, resilience and the ability to work with others – the very “basic job skills” that employers say they want.

There is also a risk that the pressure to provide placements simply transfers responsibility for training from employers to universities and students. A placement cannot compensate for employers systematically reducing entry-level recruitment. Nor can an internship replace the first job in which a graduate is trained, supported and gradually given responsibility.

The roadmap’s proposal to reduce the cost of hiring young graduates raises a related issue. The report cites employer support for extending the exemption from employer National Insurance contributions for under-21s to those under 25, while Wonkhe estimates that such a measure could cost the exchequer around £7 billion [3]. A subsidy might encourage recruitment, but it does not guarantee good jobs, quality training or a fair distribution of opportunity. It would need to be assessed against other possible uses of public money, including investment in further education, careers services and properly funded placements.

The AI problem: fewer entry points or better entry points?

The implications of AI for graduate jobs are still uncertain, but uncertainty does not mean that nothing is happening. The most immediate effects are likely to appear not as the disappearance of whole occupations but as changes in the tasks that make up those occupations.

The Stanford Institute for Economic Policy Research finds little evidence of large-scale aggregate job loss caused by AI so far. Employment in highly AI-exposed occupations has remained broadly stable, and software-developer job postings in the evidence reviewed have grown faster than postings in other occupations over the most recent period [4]. That is important: the claim that AI has already destroyed the technology labour market is not supported by the evidence.

But the same review identifies a more worrying pattern. Recent graduates are facing a difficult labour market, and there is evidence that AI may be reducing demand for young workers in some exposed occupations, including software development and customer service. Entry-level roles often contain routine research, writing, coding, documentation and analysis – the very tasks that generative AI can now perform, at least some of the time. The evidence is not conclusive because interest rates, pandemic over-hiring, remote work and wider economic weakness are also involved. Nevertheless, the possibility that AI is weakening the first rung of the career ladder should be taken seriously [4].

This is the central challenge for technology education. If junior developers, analysts, testers, technical writers and support staff are no longer hired in the same numbers, how will people acquire the experience needed to become senior developers, architects, product managers and technical leaders?

A profession cannot remain healthy if it automates away its training ground.

AI may create better jobs – but not automatically

There is also more optimistic evidence. PwC’s 2026 AI Jobs Barometer reports that the skills needed in highly AI-exposed jobs are changing more than twice as fast as in less-exposed roles. It describes a two-track labour market in which some jobs are “professionalised” by AI, requiring more judgement, leadership, creativity and strategic thinking, while other jobs are “democratised”, because AI makes tasks easier for people with less specialist expertise [5].

The report’s most striking claim is that AI-exposed junior roles are seven times more likely than less-exposed junior roles to demand traditionally senior skills such as leadership and strategic thinking. It reports growth in “seniorised” entry-level roles since 2019, even while overall early-career job postings have flatlined in highly exposed sectors [5].

If this pattern continues, universities cannot prepare graduates simply by teaching them to operate the latest AI tools. Graduates may be expected to exercise judgement much earlier in their careers, to assess AI outputs, explain decisions, work across professional boundaries and take responsibility for results produced with machine assistance.

But there is a contradiction here. Employers may want graduates to arrive with senior-level judgement while simultaneously reducing the opportunities through which junior workers develop that judgement. Asking universities to produce “work-ready” graduates is not a substitute for employers creating work in which graduates can learn.

The International Labour Organization reaches a similarly nuanced conclusion. Its refined global index estimates that one in four workers are in occupations with some exposure to generative AI, but only 3.3 per cent of global employment falls into the highest exposure category. It concludes that transformation of jobs is more likely than complete automation because most occupations consist of tasks requiring human input [6].

That is the more plausible short-term future: not a world without technology jobs, but a technology labour market in which the boundaries between coding, system design, domain knowledge, project management, communication and accountability are rearranged.

What might happen to tech jobs?

The technology sector itself is likely to divide into several different experiences.

Area of technology work	Possible AI effect	What graduates may need to demonstrate
Routine coding and software maintenance	More automation, faster production and fewer purely junior tasks in some teams	Ability to understand systems, test outputs, debug, document decisions and work with users
Data analysis and reporting	Automated extraction, visualisation and first-draft interpretation	Statistical reasoning, data quality judgement, domain knowledge and the ability to challenge plausible errors
Cybersecurity and infrastructure	AI-assisted monitoring and response alongside more complex threats	Systems thinking, risk assessment, resilience and ethical judgement
Product and service design	Faster prototyping and more synthetic research	Understanding of users, social context, accessibility, experimentation and responsibility
Technical support and implementation	More self-service automation, but greater need for integration and escalation	Communication, diagnosis, relationship management and the ability to translate between technical and non-technical users
AI governance and assurance	Expanding need to assess risk, compliance, bias, privacy and accountability	Interdisciplinary knowledge, critical reasoning and confidence working across law, policy, technology and organisations

This table is not a forecast. It is a reminder that “tech jobs” are not a single category. Some routine tasks will probably be automated. Other roles will expand because AI increases the scale of activity or creates new risks. Some occupations may be renamed while retaining much of their existing work. The boundary between technology and other sectors will become less clear as AI becomes embedded in finance, health, education, manufacturing, public services and creative work.

The ILO’s evidence also suggests that exposure is uneven. Digitised professional and technical occupations are becoming more exposed, while clerical work remains particularly vulnerable. The question is not only how many jobs disappear, but who bears the cost of transition, who gains the productivity benefits and who has access to the training needed to move into more complex work.

The curriculum should not chase the forecast

Universities UK is right that students need experience with AI and its workplace applications. But the phrase “AI trailblazer” raises a question: trailblazing what, exactly? A particular commercial platform? A set of tools that may be superseded within months? Or a deeper capacity to use technical systems critically, creatively and responsibly?

The second is more defensible. Students need to understand how AI systems work at a level appropriate to their discipline, where their outputs are unreliable, how data and bias shape results, what privacy and intellectual-property risks arise, and how to document and take responsibility for AI-assisted work. They also need opportunities to use AI in realistic professional contexts without allowing it to replace the underlying knowledge that makes professional judgement possible.

The OECD’s research on AI and work points to the importance of training and worker consultation. Workers and employers may be positive about AI’s effects on performance and working conditions, but concerns remain about job loss, data collection, privacy and algorithmic management. Training and consultation are associated with better outcomes [7]. This is relevant to universities because graduates should not be prepared merely to adapt to workplace AI. They should be prepared to participate in decisions about how AI is introduced and governed.

That means strengthening, rather than weakening, the parts of education that are most difficult to automate: disciplinary understanding, critical evaluation, communication, collaboration, creativity, ethical reasoning and the ability to learn from experience. In technology courses, this may mean placing less emphasis on producing code quickly and more on architecture, testing, explanation, user needs, security, maintenance and consequences. In non-technology courses, it may mean learning how AI changes professional practice without assuming that every student needs to become a programmer.

A better bargain between universities and employers

The Future Jobs Roadmap is most convincing when it calls for universities, employers and government to act together. It is least convincing when the graduate is treated as the main object of adjustment. If the labour market is changing, employers must also redesign entry-level work so that it remains a place for learning.

A credible bargain would have several parts. Universities would provide structured opportunities to apply knowledge, develop AI literacy and reflect on professional practice. Employers would provide paid, meaningful work-based learning and maintain genuine entry-level pathways. Government would support regional partnerships, small and medium-sized employers and accessible careers services, while also protecting young workers from being used as subsidised replacements for permanent staff.

The roadmap’s regional emphasis is valuable here. The skills required in a technology cluster, a manufacturing region, a rural economy or a public-service system will not be identical. Universities should work with colleges, local authorities, employers and community organisations to understand local opportunity—not simply to reproduce the priorities of the largest technology companies.

The first rung matters most

The impact of AI on technology jobs will not be measured only by the number of jobs eventually created or destroyed. It will also be measured by whether people can still enter the professions, develop expertise and progress into positions of responsibility.

The current evidence does not support either extreme. AI has not yet produced a general employment apocalypse, but there are credible signs of pressure on some early-career workers. At the same time, AI is increasing demand for judgement, leadership and domain knowledge in some roles. The result may be a labour market in which there are fewer routine entry points and higher expectations of those who obtain them.

That is why work-based learning matters, but also why it cannot be the whole answer. Universities should help students understand and use AI, yet employers must preserve the social process through which professional capability develops. Otherwise, the economy will demand experienced graduates while removing the jobs in which experience is acquired.

The future jobs question is therefore not simply whether universities can make graduates more employable. It is whether universities, employers and government can build a labour market in which graduates have a fair opportunity to become capable professionals—and in which AI is used to improve that process rather than quietly closing the door behind the last generation.

References

[1] Universities UK. (2026). Future Jobs Roadmap. https://future.universitiesuk.ac.uk/future-jobs/roadmap

[2] Grove, J. (2026, September 10). All undergraduates should get work-based learning, v-cs say. Times Higher Education. https://www.timeshighereducation.com/news/all-undergraduates-should-get-work-based-learning-v-c-s-say

[3] Kernohan, D. (2026, September 9). Will a graduate jobs roadmap show us the right way? Wonkhe. https://wonkhe.com/wonk-corner/will-a-graduate-jobs-roadmap-show-us-the-right-way/

[4] Stanford Institute for Economic Policy Research. (2026). What is really happening to jobs? Separating AI hype from reality. https://siepr.stanford.edu/publications/policy-brief/what-really-happening-jobs-separating-ai-hype-reality

[5] PwC. (2026, June 15). 2026 AI Jobs Barometer: Two futures for jobs in an AI era. https://www.pwc.com/gx/en/services/ai/ai-jobs-barometer.html

[6] Gmyrek, P., Berg, J., Kamiński, K., Konopczyński, F., Ładna, A., Nafradi, B., Rosłaniec, K., & Troszyński, M. (2025). Generative AI and jobs: A refined global index of occupational exposure (ILO Working Paper 140). International Labour Organization. https://doi.org/10.54394/HETP0387

[7] OECD. (n.d.). AI and work. Retrieved September 10, 2026, from https://www.oecd.org/en/topics/sub-issues/ai-and-work.html