I recently delivered my inaugutal lecture as Hon Professor of Practice. in thr School of Society and Environment at Queen Mary University of London. This is what i said
Good evening. It is a particular pleasure to be back at Queen Mary addressing what has become the defining challenge for higher education: how we prepare the workforce not just for AI-augmented work, but for a fundamentally transformed professional landscape. And can I take the opportunity at the same time to congratulate Queen Mary on its excellent showing in the Times Higher Global Rankings, particularly its 32nd place in research quality!
The question I explore throughout my book Living with the Algorithm – my attempt to make sense of AI governance and policy for a non-specialist audience – is whether artificial intelligence becomes our servant or our master. That question is nowhere more in need of resolution than in its consequences for the people sitting in our lecture theatres today.
It is a question others are asking too, at the very highest level. Only a few weeks ago, addressing the AI Summit in Scotland, echoing the Pope’s recent encyclical the King put it in his own words: that our task is to ensure AI “remains firmly in the service of humanity” – and that is exactly what is needed-not only intellectual excellence but as he said “the emotional excellence which separates man from machine.”
When I chaired the House of Lords Select Committee on AI in 2017-18, we concluded that the UK had a unique opportunity to shape AI for public benefit. More recently, I co-chaired Policy Connect’s inquiry into “Skills in the Age of AI,” which concluded last year after nine months of evidence from business, academia, and citizens themselves. What emerged reinforced for me both the urgency and the complexity of what we are discussing today.
And we now have important new data on the scale of that urgency. The third HEPI Student Generative AI Survey found that 95% of students report using AI in at least one way, and 94% say they use generative AI specifically to help with assessed work.
This is no longer a technology on the horizon. It is operating at the heart of the student experience, in this institution and every other. Any response that still treats this as something to be planned for has already been overtaken by events.
The Government’s New Framework: Promise and Limitations
The Skills England report published last October represents genuine progress. I welcome it. It identifies critical shortages – 7.3 million employed adults lacking essential digital skills, projected to become the UK’s largest deficit by 2030 – and rightly calls for “blended” professionals who combine technical knowledge with leadership and communication.
But to welcome it is not to regard it as sufficient. The report operates within the paradigm of skills acquisition – teaching people to use AI tools more effectively within existing frameworks. It does not address the structural transformation AI is driving, or grapple with how we develop system-level capabilities when the very ease of AI threatens to erode the critical thinking that underpins them.
The Employer Perspective: What’s Actually Valued
Recent research from Kingston University provides stark evidence of the gap between what employers need and what graduates provide. Their latest Future Skills report reveals that 56% of businesses are now likely to consider skills-based recruitment – a fundamental shift from credential-based to capability-based hiring. Kingston identified nine core future skills employers value, from creative problem-solving to a questioning mindset to resilience. Notice what dominates that list: digital competency is one of nine. The rest are adaptive capabilities.
The World Economic Forum’s 2025 Future of Jobs Report identifies analytical thinking, creative thinking, resilience, flexibility and agility as its top skills. Yet Kingston’s research reveals a concerning disconnect: only 23% of UK businesses anticipate AI will fundamentally change their business model in the next five years, compared with East Asian nations treating this as a top strategic priority. On this measure, the UK risks lagging behind.
The Friction Paradox: Why Ease Threatens Capability
Here is the challenge that I find most striking, and most underappreciated. The emerging evidence is that AI’s ease may be undermining the very critical thinking we are trying to develop.
The cognitive psychologist Ronald Kellogg spent years studying how people write before concluding that drafting a basic persuasive essay is as mentally taxing as digging a ditch is physically exhausting. I thought this audience would appreciate that analogy!
That difficulty is not incidental – it is the mechanism by which writing works. Kellogg’s own phrase for it is one I want to leave with you this evening: writing is a technology for thinking. Any tool that removes the effort risks removing the thinking along with it.
Daniel Susskind’s excellent new book What Should My Children Do? is, by design, an optimistic one: he wants us to prepare children for the world they will actually inhabit – the water, as he puts it, in which the next generation will swim – and he is confident that personalised AI tutors and a reorganised curriculum can help them flourish in it.
But he is not naïve about the discipline required. If AI makes education too easy, he argues, the burden shifts to us to make it more testing – to push young people on and stretch them – and he insists that we use AI critically rather than blindly, keeping our basic skills sharp. Where we differ is in emphasis. His optimism is about the destination; mine is about the discipline needed to reach it safely.
A 2025 study found a significant negative correlation between frequent AI use and critical-thinking performance, especially among younger users – the mechanism researchers call “cognitive offloading”: when AI provides effortless answers, users stop doing the evaluative reasoning that builds judgment. This connects to Daniel Kahneman’s framework of thinking fast and slow. AI encourages System 1 thinking – fast, fluent, and instinctively trusted – while genuine learning requires the slower, effortful System 2 engagement that questions assumptions. Productive friction is not pedagogical stubbornness; it is the necessary intervention that forces that engagement.
It is striking that Joseph Stiglitz has recently drawn a very similar distinction from an entirely different direction. He separates what he calls AI from what he prefers to call IA – intelligence assisting – reaching for the same analogy scientists have long used for the microscope and the telescope: instruments that extend what we can see without ever doing the seeing for us. In his own research, Stiglitz is emphatic that AI is “not a substitute for thinking.”
An MIT Media Lab experiment found that students who relied on ChatGPT performed worse, remembered less, and were less cognitively engaged than peers who wrote without AI – recording lower brain activity, struggling to recall what they had written, and reporting little sense of ownership over the work.
Compelling corroboration has emerged partly here at Queen Mary, where researchers studying medical students’ diagnostic decisions found that those using AI alone scored lowest – yet were the most confident, despite being the least skilled. One researcher likened it to giving learners a powerful sports car before they have learned to drive. The lesson is unambiguous: AI is transformative only when used at the right time, with the right training, and under strong governance.
A large-scale study of 26,811 pupils in China, cited in a major new UK report on AI in education published last month, found the same pattern at national scale: AI tools raised homework scores by 18%, while performance on unassisted exams fell by 20%. The help is real. So, unmistakably, is the cost.
Beyond Skills to Readiness
Our Policy Connect inquiry revealed a troubling gap. Despite massive investment in AI training, a 2024 survey found only 21% of workers feel “very confident” integrating AI into their workflows, and McKinsey found that while 89% of organisations use AI, only 9% have achieved “AI maturity.” The HEPI survey makes the institutional dimension concrete: while 68% of students believe generative AI skills are essential, fewer than half feel their teaching staff are helping them develop those skills – a twenty-point gap that is widest among Arts and Humanities students, the cohort most exposed to AI’s disruptive effects and least likely to feel prepared for what is coming.
Three Emerging Patterns
Research from the Centre for Finance, Technology and Entrepreneurship (CFTE) identifies three patterns defining graduates’ trajectories. The first is Mass Displacement – roles built around execution being steadily automated. It often begins quietly, with shrinking scope and fewer chances to grow, and it is gradual until it becomes sudden. Entry-level graduate roles are among the first places it shows: a 2025 Harvard study found sharp declines in junior-level hiring while senior hiring remained flat, and the National Foundation for Educational Research projects up to three million UK jobs in declining occupations could disappear by 2035.
This is not speculative; it is measurable, happening now – and increasingly, the remaining positions are filtered by AI recruitment systems before a human ever sees the application: so ironically we are preparing students for careers where AI decides if they are qualified to work with AI.
The second pattern is the Supercharged Professional – individuals achieving ten to a hundred times their previous output, restructuring how they work rather than merely learning new tools.
The third is the Creative Disruptor, building entirely new systems.
None of this is predetermined. Displacement is not inevitable – but it is where we will end up if nobody acts. And I would argue that universities are where that action must begin.
The Performance Hexagon
CFTE describes five patterns of contribution, which it calls the Performance Hexagon. I will focus on the top four. Task Robots do the job reliably, as instructed. Problem Solvers work things out for themselves. System Thinkers design structures that solve whole categories of problems. Superstars spot opportunities that no one has pointed out. I should emphasise that these are patterns, not labels – the same person can be a Problem Solver in one job and a Task Robot in another.
Now add AI. At the lower levels, where work is routine, AI replaces. From Problem Solver upwards, it amplifies. CFTE’s point is that learning to use AI tools only moves you faster along the same line. What protects you is moving up. So the question for us is whether we are preparing graduates to climb – to build the questioning mindset and adaptability that separate those who are supercharged from those who are displaced. That is the gap our universities must fill.
Hard Hats and Graduation Caps
I want to address one further strand of the current policy conversation, because Queen Mary sits in an unusually good position to tackle it. The new Prime Minister has made technical education a defining project of his government, telling students this summer that Britain will now “value the hard hat every bit as much as the graduation cap.” I welcome the underlying instinct – Alan Milburn’s warning of a generation being left behind is a genuine crisis, and Queen Mary’s own substantial investment in graduate and degree apprenticeships puts this institution ahead of most of the sector in taking that parity seriously in practice.
But I want to enter one note of caution about how that parity is being sold. Some of the enthusiasm for the hard hat rests on an assumption that trades are somehow AI-proof in a way graduate careers are not. I do not think that assumption will survive contact with the technology – AI-assisted diagnostics and scheduling tools are already moving into construction and skilled trades, just as equivalent tools are moving into law and journalism.
Writing recently, Graham Watts of the Construction Industry Council and Aled Williams of the University of the Built Environment make much the same point – that parity between technical and academic education is right, but it “cannot become a binary choice between the hard hat and the graduation cap,” since the built environment needs people across a whole spectrum of blurring roles.
Mark Farmer of the Construction Leadership Council goes further: young people are already pivoting toward roles they perceive as “AI-proof” regardless of government messaging, but push the pendulum too far and apprenticeships themselves become oversubscribed relative to available jobs – with under 27% of plumbing apprenticeships even completed, and 700,000 graduates already jobless, neither pathway as currently designed is working well.
Daniel Susskind’s book, What Should My Children Do?, offers a cautionary precedent for exactly this risk. He reminds us that the coalition government once rushed to put coding at the centre of the curriculum, confident it was future-proofing a generation. Coding turned out to be one of the things AI does best, and those skills were largely redundant before the children concerned had even left school. His conclusion is that we should resist the promise of “future-proofing” altogether, because we know too little about what lies ahead to bet a curriculum on any single skill. If a government could get this wrong once, with software, it can get it wrong again, with bricklaying or plumbing recast as this decade’s safe bet.
The Prime Minister is right to elevate technical education. But respecting a path means teaching it as seriously as any other. If these new courses just teach tools, they’ll have the same problem as universities — teaching people to use AI, not to think. What matters isn’t the hat you’re wearing. It’s whether you’ve been taught to think, or just to follow instructions.
What Readiness Demands
The fundamental divide is who does the thinking. That is CFTE’s phrase, and it sits at the heart of their argument: even a very good Task Robot has someone else doing the thinking for them, while Problem Solvers, System Thinkers and Superstars define the problem for themselves.
CFTE’s answer is that future-ready people need three things, and I draw here on their three pillars. Domain expertise: the knowledge of how an industry actually functions — something AI cannot replace. Technology fluency: the ability to structure a workflow around AI and judge what it produces. And adaptive capability: the discipline to think clearly, solve problems independently, and operate without a script.
CFTE also notes something that should concern us here: some younger professionals are fluent with the tools but lack the domain expertise to judge what the tools produce. It echoes what the Queen Mary medical study found.
Deloitte found that 66% of managers believe recent hires are unprepared, identifying an “experience gap” larger than the skills gap: what is missing is judgment and contextual understanding, not technical knowledge. The World Economic Forum projects 44% of workers’ core skills will change within five years. Technical skills alone provide insufficient protection.
The case for the value of a degree doesn’t only come from universities. It comes from inside the technology industry too. Valerie Capers Workman, formerly vice-president of people at Tesla – working for one of technology’s loudest sceptics of university education – has this year argued the opposite case: that a degree is more valuable today than ever, precisely because the ability to reason, question, and lead with humanity will be sought after in a way machines cannot replicate, singling out the liberal arts as an area of growing value.
But who does the thinking is not only a pedagogical question. It is a distributional one. Joseph Stiglitz has argued that AI risks becoming a textbook case of a general-purpose technology turbocharging inequality – stripping labour out of production, concentrating profits at the top, and pushing the costs of transition onto workers and the public. Graduates who leave with genuine judgment are better placed to sit on the right side of that divide than those who leave only with tool proficiency, however fluent. What we choose to teach is a matter of who, in a decade’s time, has any bargaining power at all.
The Transformation Required
Our Policy Connect inquiry recommended making AI literacy mandatory in the National Curriculum. For universities, the implications are equally profound.
First, AI literacy must be embedded across all disciplines – a law graduate needs to understand algorithmic decision-making as thoroughly as a computer science graduate needs data protection, with assessments where students critique AI outputs and reconcile them with primary sources.
As the educator Nick Potkalitsky puts it, the goal is not teaching which button to press, but shifting from “does this sound right?” to “how would I check this?” – and that shift cannot be achieved through a single professional development session; it requires embedded practice and ongoing support.
Second, we must teach judgment. The Post Office Horizon scandal shows the catastrophic cost when professionals cannot challenge automated systems. Our graduates need to understand not just using AI in diagnosis, but ensuring accountability; not just deploying algorithms, but auditing them for bias.
Third, how we assess students must evolve. The HEPI survey found 65% of students say assessment has already changed significantly in response to AI, up from just 32% two years ago – a trajectory that will continue whether or not institutions shape it deliberately.
We do not have to invent this from scratch. The University of Sydney has adopted a two-lane approach: every discipline sets some assessments in person, with AI restricted, and others that require AI, so students demonstrate they can apply it well. That takes seriously both halves of what employers want – independent judgment and technological fluency – rather than treating them as a single sliding scale.
Looking at the history of the electronic calculator, Daniel Susskind arrives at a very similar prescription: neither ban these tools nor cling to inherited ways of working, but teach children to operate competently both with AI and without it, while doubling down on the basic skills that should never erode. A major new report last month, Sir Anthony Seldon and Tim Bunting’s Britain’s Greatest Education Opportunity Ever, goes further still, recommending that government adopt exactly this two-lane model nationally – reserving secure, supervised assessment for a small number of high-stakes moments, and opening the rest to AI-supported, continuous evaluation. That they have all arrived, from different starting points, at the same dual-track design, is about as strong a signal as we are likely to get that this is the right direction for the years ahead.
Enabling Structural Change
Skills England identifies that employers want “bolt-on” training – short, modular options rather than multi-year apprenticeships – which signals that current structures are too rigid. Proposals for lifelong skills grants, or replacing the apprenticeship levy with a more flexible skills and training levy, would fund exactly this. Our inquiry’s recommendation to relaunch Local Digital Skills Partnerships – which, on a modest £75,000 per region, upskilled 12,000 workers and reduced digital exclusion by 18% – shows the model can work again.
Building Trust Through Transparency
Throughout my own work on AI regulation, I have argued that public trust is foundational to everything else. When our inquiry took evidence, only 33 public sector algorithmic tools had published transparency records. Today the government’s own register lists 152, and I welcome that progress. But the standard is mandatory only for central government and its arm’s-length bodies. Elsewhere it is voluntary, and take-up is thin: the register shows only about ten councils and two police forces. Progress, then – but not yet a foundation for trust.
This isn’t just a design principle. It’s a warning. Stiglitz points out that the same people pushing hardest for fast, unmanaged AI adoption are also campaigning to shrink the government capacity we’d need to manage the fallout. He draws a telling parallel: when machines displaced farm workers in America, it took decades – and a war – before government finally moved them into new jobs. Left alone, adjustment doesn’t happen. Someone has to build it. Universities are one of the institutions that can. Our job isn’t to resist oversight of these technologies. It’s to produce graduates who can design it, demand it, and work within it.
Digital Exclusion
Our Policy Connect inquiry found 19 million people in the UK face digital poverty. The HEPI survey adds a dimension that should concern universities specifically: students from higher socioeconomic households are measurably more likely to be using advanced AI tools than those from lower socioeconomic backgrounds – a gap that persists even where free versions are available, meaning cost alone does not explain it.
The same divide shows up starkly at the other end of the system: only 20% of young people not in education, employment or training have ever used an AI tool at all, against 93% of postgraduates – precisely the group Alan Milburn has already warned who risk losing out altogether. Providing access to AI tools is not enough; the skills divide is becoming a new dimension of educational inequality, and it is forming now, among our current students and among those who never make it into a classroom at all.
Conclusion
Based on the rather speculative figures provided in the AI Opportunities Action Plan of 2025, Skills England estimates AI adoption could boost the UK economy by £400 billion by 2030 – but even that assumes we get skills development right.
Queen Mary has always combined academic excellence with social purpose, and the research emerging from this institution is helping build the evidence base that should inform how every university in this country approaches these questions. The UK can lead in AI – not through fastest adoption or lightest regulation, but through the most thoughtful, ethical, human-centred approach. That leadership must begin in universities, and it should begin now.
But we must be honest about the depth of change required. Adding AI modules will not suffice. Providing unfettered AI access without pedagogical friction will not suffice. We need to fundamentally rethink professional capability development in an age where AI’s ease paradoxically threatens the cognitive capacities we are trying to develop.
Peter Hyman, addressing the Royal Society of Arts only a few weeks ago, called for what he termed a “human-maxxing” revolution in education – developing everything young people are capable of, head, heart and hand, and taking back control of their attention, their judgment, and their futures. I can think of no better description of what I have been arguing for this evening.
Thank you.
Endnotes:
Government & Policy Documents
Skills England. (2025). AI Skills for the UK Workforce. Department for Education. https://www.gov.uk/government/publications/ai-skills-for-the-uk-workforce
HM Government / Department for Science, Innovation and Technology. (2025). AI Opportunities Action Plan. [Original source of the estimate that AI adoption could boost the UK economy by up to £400 billion by 2030; repeated in the Skills England report above.]
Policy Connect. (2025). Skills in the Age of AI. Co-chaired by Lord Clement-Jones and Lord Holmes. https://www.policyconnect.org.uk/research/skills-in-the-age-of-ai
Seldon, A. & Bunting, T. (2026). Britain’s Greatest Education Opportunity Ever. [Source of the 26,811-pupil China study, the recommendation to adopt the two-lane assessment model nationally, and the 20% (NEET) vs 93% (postgraduate) AI-use comparison.]
The King. (2026, 17 September). Speech at the AI Summit in Scotland, Dumfries House. The Royal Family (royal.uk). [Source of the quotations in the opening.]
Prime Minister’s Office / GOV.UK. (2026). Press release: “PM: From today Britain will value the hard hat as much as the graduation cap.” [Source of the direct quotation in “Hard Hats and Graduation Caps.”]
Lloyds Banking Group. (2024). 2024 Consumer Digital Index. https://www.lloydsbank.com/consumer-digital-index.html — NB: please confirm this is the intended source for the “19 million people… face digital poverty” figure. Lloyds’ own published figures do not clearly total 19 million in any single year checked; the Digital Poverty Alliance’s overlapping estimate of 13-19 million may be the source actually used by the inquiry.
Department for Science, Innovation and Technology / Government Digital Service. (2026). Algorithmic Transparency Records (GOV.UK repository; 152 records as at 10 September 2026).
University Research
Kingston University. (2025). Future Skills 2025: Perspectives from East and West. In partnership with Nanyang Technological University, Singapore. https://assets.kingston.ac.uk/m/32c8151479564c48/original/20250617_FutureSkills2025_report-1.pdf
HEPI / Kortext. (2026). Student Generative AI Survey 2026. Report 199. Rose Stephenson and Charlotte Armstrong. Higher Education Policy Institute.
Academic Research — Cognitive Science & AI Impact
Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15, Article 6.
Kahneman, D. (2011). Thinking, Fast and Slow. New York: Farrar, Straus and Giroux. ISBN: 978-0374533557
Kaplan, A. (2025). Productive Struggle: The Imperative of Friction in AI-Driven Learning.
Kentz, M. & Potkalitsky, N. (2024). AI in Education: A Roadmap for Teacher-Led Transformation. ISBN: 9780970357144
Potkalitsky, N. (2025). The Humanist Librarian in the Age of Generative AI.
Queen Mary University of London / Erasmus University Rotterdam / University of Campinas / McMaster University research on AI-assisted diagnostic learning in medical education (cited in HEPI Report 199, 2026).
Kosmyna, N., Hauptmann, E., Yuan, Y.T., Situ, J., Liao, X-H., Beresnitzky, A.V., Braunstein, I., & Maes, P. (2025). “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task.” MIT Media Lab. arXiv:2506.08872. [Corrected citation – this, not the MIT Ad Hoc Committee report, is the source of the brain-activity, recall, and ownership findings cited in the Friction Paradox section.]
Stiglitz, J. (2026, 31 May). Interview: on AI, inequality, and “Intelligence Assisting” (IA). Fortune. [Corrected date – previously misdated 2025.]
Susskind, D. (2026). What Should My Children Do? How to Flourish in the Age of AI. London: Allen Lane (Penguin).
Press Coverage
Watts, G. (2026, 16 August). “Both mortar boards and hard hats can lead to work security.” The Observer. [Includes quotations from Aled Williams, University of the Built Environment, and Mark Farmer, Construction Leadership Council.]
Clarence-Smith, L. (2026, 2 September). “University does equip you for world of AI, whatever tech leaders say.” The Times. [Source of the Valerie Capers Workman quotation.]
Woolcock, N. & Lambert, G. (2026, 22 September). “Schools must teach pupils to ‘think for themselves’ in AI-driven world.” The Times. [Source of Peter Hyman’s “human-maxxing” remarks at the Royal Society of Arts.]
Susskind, D. (2026, 6 September). “I’m a father of three who studies the impact of artificial intelligence: this is what parents need to know about AI.” The Guardian. [Supporting source for the coding precedent, “teach both, test both,” and the case for basic skills.]
Susskind, D. (2026, 10 September). “How to teach the next generation in the era of AI.” Financial Times.
Usborne, S. (2026, 9 September). “What Should My Children Do? by Daniel Susskind review – the case for embracing tech in the classroom.” The Guardian.
International Organizations
World Economic Forum. (2025). The Future of Jobs Report 2025. https://www.weforum.org/publications/the-future-of-jobs-report-2025/
The Holy See: Encyclical Letter Magnifica Humanitas of His Holiness
Pope Leo XIV on Safeguarding the Human Person in the Time of Artificial Intelligence: https://www.vatican.va/content/leo-xiv/en/encyclicals/documents/20260515-magnifica-humanitas.html
Management Consulting Research
McKinsey & Company. (2025). The State of AI in 2025.
Deloitte. (2025). The Skills Gap in Focus.
Education & Workforce Research
Centre for Finance, Technology and Entrepreneurship (CFTE). (2025). The AI-fication of Talents Whitepaper: Why some individuals, organisations and nations are powering ahead while others are falling behind. Version 1.0, June 2025. [Source of the Performance Hexagon, the three pillars of AI-readiness, the three patterns, and the “who does the thinking” framing.]
Nguyen Trieu, H. (2024). The AI-fication of Jobs. CFTE. [Origin of the three patterns: mass displacement, supercharged professionals and creative disruptors.]
EngageEmployee. (2024). Employee survey on AI confidence, as cited in CFTE (2025). [Source of the 21% figure.]
Harvard Business School. (2025). The Impact of GenAI on Entry-Level Employment.
National Foundation for Educational Research. (2025). Skills Imperative 2035.
Parliamentary
House of Lords Select Committee on Artificial Intelligence. (2018). AI in the UK: ready, willing and able? HL Paper 100.
https://publications.parliament.uk/pa/ld201719/ldselect/ldai/100/100.pdf
25th April 2026
Lord C-J on AI and the Future of Work
27th November 2022
