Executive summary. Artificial intelligence is reshaping skills, education, and the nature of work across every sector and economy. As AI transforms what organisations and society need, the capabilities it cannot reliably replicate—judgement, adaptability, critical thinking, effective collaboration—are becoming more important. Yet globally, only 32% of employers believe education is developing these capabilities adequately. Proven models exist to change this, but they operate in isolation. Closing the gap requires coordinated action from educators, employers, and policymakers working together.
Key takeaways
- AI amplifies the value of human-centric capabilities while making them harder to develop — and the quality of human reasoning now determines the quality of AI output, making this a technical imperative as much as a humanistic one.
- Effective models share a consistent architecture: skills embedded throughout the learning journey, learning built around authentic, real-world challenges, and credentials that travel with the graduate.
- The coordination infrastructure to connect what works — shared frameworks, accessible practice pathways, and feedback mechanisms that operate at the pace of labour market change — does not yet exist at scale. Building it requires educators, employers, and policymakers to act together.
The Question Science Fiction Answered First
For decades, film and literature have grappled with the issue of what machines cannot do. The androids in Blade Runner lack the capacity for moral reasoning. HAL 9000 in 2001: A Space Odyssey executes instructions but cannot navigate ethical ambiguity. Ava in Ex Machina can simulate human emotion, but not sustain genuine empathy. These portrayals were, in hindsight, foreshadowing: skills like collaboration, ethical decision-making, and contextual judgement are and will remain uniquely human.
That insight has moved from fiction to research consensus, and the implications for education are now pressing. AI is restructuring the nature of work in real time, accelerating the pace at which skills requirements change, and exposing the gaps between what education produces and what employers and economies actually need.
The Data Behind the Disconnect
The World Economic Forum’s 2025 research is direct: while generative AI can transform technical tasks such as coding and data analysis, it has minimal impact on human-centric capabilities.4 While these skills are not diminishing, they are more difficult to develop. The research shows that these skills cannot be acquired through lectures or passive digital exposure. They emerge through deliberate practice in authentic contexts, with real feedback, over sustained periods.
The most recent global evidence comes from a 2026 study across six countries commissioned by Pearson and AWS.1 Only 24% believe universities are currently keeping up with the pace of market developments. An even starker finding is a 46 percentage point divergence between HE leaders’ belief that their graduates meet employer expectations contrasted to what employers believe. A further gap exists in the opposite direction: HE leaders consistently undervalue communication, collaboration, and adaptability relative to what employers actually demand.1
The pattern is consistent across data sources. The QS Global Employer Survey, drawing on thousands of employers worldwide, identifies the same five gaps—critical thinking, problem-solving, communication, adaptability, and emotional intelligence—as the capabilities most consistently missing in graduate hires.2 A further dimension comes from BetterUp’s longitudinal research: human-centric capabilities declined sharply between 2019 and 2021 and have not recovered, creating what researchers describe as a ‘capability recession’.3 The gap is widening.
Why AI Deepens Rather than Resolves the Gap
Higher education has responded to the AI moment largely by adding ethics content to existing curricula and developing policies on AI use in assessment. The Pearson/AWS study found that while 47% of institutions report ethics is widely covered, working productively alongside AI is the least-taught topic. Meanwhile, employers rate graduates’ ability to critically verify AI outputs as their weakest competency, which illustrates that awareness without practice does not build judgment or AI fluency.
Three strands of research clarify what is actually at stake. First, Anthropic’s research on AI fluency finds that the value generated through AI is determined by the quality of human reasoning applied to it.5 Constructing meaningful prompts, interrogating outputs critically, and iterating based on a nuanced understanding – these are human skills. Weak human reasoning produces weak AI outcomes; strong human skills are what make AI genuinely useful.
Second, the Alan Turing Institute’s “Doing AI Differently” research (July 2025) makes a related argument at the level of AI design itself.6 The initiative—involving 150+ researchers across six continents—finds that humanities skills, contextual judgement, and the capacity to reason about meaning make AI systems more effective and more trustworthy.
Third, as AI takes on more consequential roles in organisations—in hiring, lending, healthcare, and resource allocation—the ability to interrogate AI recommendations, identify where bias or error may have entered, and take ownership of decisions that affect others becomes not just a professional skill but an ethical one. The Pearson/AWS study frames this directly: graduates are expected to serve as an’“ethical filter’ for employers, responsible for safety, integrity, and the responsible deployment of technology.1
Meanwhile, the practice pathways that build these capabilities are narrowing. Entry-level roles are changing in nature: WEF research shows the shift is away from fixed processes and towards applying judgement, creativity, and collaboration alongside technology.7 At the same time, the informal routes through which previous generations built these skills—part-time work, Saturday jobs, in-person internships—are shrinking. Students are being asked to demonstrate more sophisticated capabilities at the same time as opportunities to practise them are disappearing.
What Works—and the Architecture Behind It
Across continents, models demonstrate that human-centric skills can be developed intentionally and at scale. What distinguishes effective approaches is a consistent underlying architecture: skills embedded throughout the learning journey; learning built around authentic challenges; assessment through demonstrated performance; and a deliberate culture of learning and working critically and responsibly alongside AI.
At the institutional level, among the clearest examples is the University of East London that integrates development experiences progressively across each year of study, covering digital proficiency, cognitive skills, emotional intelligence, ethics and cultural competency, and social intelligence, with work-based learning built in at each stage.8 UCL’s Industry Exchange Network makes live employer challenges the actual curriculum: companies submit real problems and students work on them alongside both an industry mentor and an academic supervisor.10 Illinois Institute of Technology has gone further still, adopting an explicit institutional philosophy it calls “Humans-in-the-Lead”: orienting its entire curriculum around graduates who use AI as input to decisions they own.9 This is built around what the university describes as a trifecta—technical and data skills, combined with critical thinking, systems thinking, and ethical reasoning—all at equal curricular priority, with an AI Plus module embedded across every degree programme. As its President, Raj Echambadi, puts it: “You can’t out-compete technology by giving students more technology education. What we’re finding from employers is that qualitative judgement and decision-making under uncertainty is becoming far more important.”9
At the sector and national level, intermediary platforms and government-coordinated systems remove the dependency on geography and personal networks. Parker Dewey in the US partners with 5,900+ institutions, matching students to short professional projects that require genuine judgement.11 Junior Achievement Worldwide reaches 20 million young people annually across 100+ countries, giving students consequential practice in negotiation and decision-making without career risk.12 Singapore’s SkillsFuture connects skills frameworks to funding incentives and structured pathways;13 Canada’s ICTC Work-Integrated Learning programme co-funds 50–70% of student wages in digital placements, making participation viable for employers of all sizes.14
These models are not directly transferable—governance structures and funding mechanisms differ across systems—but they demonstrate what works and that coordination at scale is achievable when the roles of universities, employers, and governments are clearly defined.
Why the Coordination Gap Persists and How to Close It
The human skills gap is not primarily an awareness problem or a curriculum design problem. The evidence, the models, and the architecture for action are available. What is missing is the infrastructure to connect them at scale: shared frameworks for what graduates should be able to demonstrate; pathways that give more students access to practice regardless of background or location; and feedback mechanisms that operate at the pace of labour market change rather than curriculum review cycles.
Institutions compete on metrics—research output, student retention, league table rankings—that do not reward sharing infrastructure or embedding employer perspectives in curriculum design. Assessment traditions favour what can be easily standardised. Smaller employers, who hire the majority of graduates in most economies, are largely excluded from existing partnership models. And the experiences students do gain rarely translate into portable, recognised evidence of capability.
No single actor can close this gap. Universities bring pedagogical expertise and the reach to develop these capabilities at scale. Employers bring real-world context and the ability to signal what matters. Policymakers can create the incentive structures and shared infrastructure that make coordination economically viable for all parties. The ability for all three to act together will shape both equity outcomes and economic competitiveness in the AI era.
How well do the systems we design—for learning, for hiring, for professional development—actually reflect what we now know about how human capabilities are built? And what would need to change—in incentives, in institutional cultures, or in the relationships between education and work—for that to happen at scale?
References
1 PSB Insights (2026). AI Readiness: Building the Bridge from Higher Education to Work.
Commissioned by Pearson and AWS.
2 Best, A. (2026). The Skills Universities Need to Build in Their Graduates Now. QS Quacquarelli Symonds.
3 Niederhoffer, K. (2026). Why the Workforce Is Less Capable Than It Was in 2019. BetterUp / World Economic Forum.
4 World Economic Forum (2025). New Economy Skills: Unlocking the Human Advantage.
5 Anthropic (2025). The AI Fluency Index: Measuring Human–AI Collaboration
6 Hemment, D., Kommers, C., et al. (2025). Doing AI Differently: Rethinking the Foundations of AI via the Humanities. White Paper. The Alan Turing Institute
7 World Economic Forum. Learning-to-Earning Sandbox website
8 Career Zone, University of East London (2026). Graduate Attributes Framework: Embedding Human-Centric Skills for Graduate Employability. Presentation at the Cappfinity Skills Symposium, Woburn House, London, April 2026.
9 PSB Insights (2026). AI Readiness: Building the Bridge from Higher Education to Work. Case study: “Humans-in-the-Lead” Framework, Illinois Institute of Technology.
10 PSB Insights (2026). AI Readiness: Building the Bridge from Higher Education to Work. Case study: Industry Exchange Network, University College London.
11 Parker Dewey. Micro-Internship Platform and Outcomes. https://www.parkerdewey.com
12 Junior Achievement Worldwide. Programme Information and Global Impact.
13 SkillsFuture Singapore. National Skills Framework.
14 Information and Communications Technology Council (ICTC) Canada. Work-Integrated Learning Digital Programme.
UK-specific sources: Refs 8, 10. US source: Ref 9. Global sources: all others. The Turing Institute initiative (Ref 6) is UK-based but draws on a research community spanning six continents.
Anila DeHart is an Adviser and Researcher on Education–Workforce Coordination and an Executive Facilitator at Headspring, a joint venture between the Financial Times and IE Business School. She is a former Global Managing Director at Deloitte. She advises education institutions on curriculum portfolio alignment for the future of work and works with organisations to redesign early career programmes to meet the demands of an AI-era workforce. She is collaborating with the Higher Education Policy Institute (HEPI) on research into university–employer partnership infrastructure and contributes to the University of Oxford AIEOU shared research agenda on the future of education, work, and institutions.
