It’s hard to keep track of the statistics predicting the profound transformation AI will bring about in the jobs market. But what is becoming clear is that employers are seeing more than a skills gap. It is a structural mismatch between how the future workforce is educated and reality.
Higher education is built to produce human performance within set assumptions: a relatively stable job market; predictable competencies; curricula designed for a world that evolved slowly. For decades, it prepared students for fair-weather racing, optimising sails and navigation within a known course, clear rules, predictable conditions. This was rational. It worked.
AI is the perfection of that model. It learns by mimicry to perform at scale. It will soon execute within defined parameters faster and more reliably than 99% of graduates in cognitive tasks. At the same time, the conditions have changed. The course towards career success is no longer marked. Jobs, tasks, and required skills are shifting faster than any programme can track. What the world now demands goes beyond fair-weather racing. It is more about open-sea navigation: reading the weather, interpreting signals that have no manual, and sometimes discovering the destination only by sailing toward it.
Current trends impacting AI in education
Faced with uncertainty, the instinct of students and their families is to run toward the most usable certifications. Business-related degrees are the fastest-growing major field in the US, with 1.63 million undergraduates enrolled in 2025. In the UK, Business and Management now accounts for 21% of all university enrolments. The logic is understandable. But the data is not encouraging. In the US, around 44% of recent business graduates are underemployed. Doubling down on applied credentials would be a rational response in a stable job market. In this one, it is preparation for a world that is already disappearing.
Trends from researchers and the expert community points in a better direction. The World Economic Forum’s Future of Jobs Report 2025 identifies socio-emotional skills, such as empathy, social influence, resilience, communication and collaboration, are among the fastest-rising competencies employers seek. As AI absorbs increasingly complex cognitive tasks, the professionals who thrive will be those who bring these capacities to their work. This consensus is right and important. How higher education delivers on it is crucial but still a challenge ahead.
The real value of higher education
There is another gap that is less talked about. It concerns the quality of thinking itself in this world that is becoming more and more uncertain. With AI excelling at performing within well-defined parameters, navigating uncertainty is what human education must now deliberately train for.
Meanwhile, in the past decades, higher education has progressively built its intellectual model around the opposite.
It can and should begin with wonder: the moment something arrests you, not because it breaks a pattern, but because the pattern itself seems wrong: insufficient, unjust, or simply not inevitable. Aristotle called it thaumazein, the origin of philosophy. Modern education has nearly eliminated it. Students today are trained to solve problems handed to them, pre-formed and bounded. The capacity to notice that the idea could be otherwise, to be unsettled by what everyone else has accepted, is rarely cultivated beyond intellectual dissertation. AI can detect anomalies within a statistical distribution. It cannot question whether the distribution itself should exist.
From wonder comes interpretation: making sense of the tensions, of what is missing, of what has not yet been articulated. A good lawyer reads a contract for what it does not say. A strategist reads a market for the signal obscured by noise. This requires a point of view shaped not just by knowledge but by inhabiting the consequences of one’s judgments. The relevant distinction is not technical: it is that AI does not live inside the outcomes its outputs produce. It does not carry them forward. Only very few institutions — Oxbridge, some liberal arts colleges — have preserved substantial space for this kind of undergraduate inquiry. It also requires a plurality of mental models: understanding that consumption patterns or voting behaviour are tied to psychology, sociology, politics and history simultaneously, not to any single discipline.
Interpretation enables conception of new ideas: imagining what does not yet exist. Holding contradictory constraints together long enough to produce something new from the tension between them. This is what designers do, what legislators do, what founders do — giving form to new products, services, relationships, regulations, or masterpieces. Conception in this context requires judgment: the capacity to decide what is worth building when no metric compels the choice, to commit to a direction that forecloses other directions, to hold responsibility for the outcome. What do I want to build? That question should be the spine of an education.
All this takes us back to the roots of what education is about: wonder, sense-making, imagination, judgment and ethics. These are not soft additions to a technical curriculum. They are the preconditions for using powerful tools well — and for asking the questions that no tool will generate on its own. Once we truly acknowledge uncertainty as the new paradigm, and AI as its sharpest accelerant, making these capacities the deliberate core of education is not a nostalgic gesture. It is the only serious response.
