Return on Education: The Future Education Bargain
On AI, credentials, bildung, and what schools might quietly need to become.
For most of the last hundred years, education in Denmark and much of the rest of the world ran on a bargain that was simple enough to fit on a postcard. You studied, you qualified, you entered society with a recognisable role, and the institution that handed you the certificate vouched for your readiness to inhabit it. The bargain was never as clean as the postcard suggested — class, capital, family and luck always tilted the field — but it was coherent enough to organise an entire life around. Parents understood it. Employers understood it. Students could see a path from where they were standing to where they were trying to go, even when they did not particularly enjoy walking it.
Photo by Jaredd Craig on Unsplash
That bargain is fraying now, and not in some distant or speculative way. In 2026, the vice-chancellor of King's College London was already saying out loud that a degree is no longer a guaranteed passport to social mobility in an era of mass higher education and a graduate job market under quiet pressure from AI. Faculty surveys in the United States showed ninety per cent of college teachers expecting AI to erode their students' critical thinking, and ninety-five per cent expecting deeper technological dependency. Denmark, meanwhile, is on track for a shortage of roughly a hundred thousand skilled workers by 2030, while continuing to send around half of its thirty- to thirty-four-year-olds through university programmes that the labour market does not always know how to absorb. Each of these is its own story. Read together, they describe a system that has stopped agreeing with itself on what it is meant to do.
What makes this moment interesting, rather than simply alarming, is that it is forcing a question that education has been able to postpone for most of the last century. Not the technical question of curriculum, or even the institutional question of funding and structure, but the older and harder question of what education is actually for once a certificate stops doing the work it used to do.
A system pulled in eight different directions
The most useful way I have found to read the current moment is to stop looking for a single trend and start watching the signals interfere with one another. At ANTICIPATE we have spent the last couple of years working closely on these questions across a fairly broad set of contexts — universities trying to redesign their programmes, NGOs working with young people on futures literacy, public sector bodies looking at lifelong learning, and several rounds of foresight training where teachers themselves get to sit in the participant chair. (The pattern that keeps showing up is not really a curriculum problem. It is a series of unresolved tensions that the system used to be able to hold together and is increasingly struggling to.
The first is the slow split between STEM and the humanities under the pressure of AI. In engineering, computer science, and the harder sciences, generative AI is being absorbed as a research instrument relatively quickly, the way calculators and statistical software once were. In the humanities, the same tools are doing something quite different to the underlying activity. When the writing, reading, and argumentation are the discipline, outsourcing them to a model is not a productivity gain — it is a quiet evacuation of the muscle the discipline was meant to build. The result, increasingly visible in faculty rooms and in student work, is that two halves of the university are walking away from one another at a speed nobody quite planned for.
The second is the credentials-and-capabilities mismatch playing out at the labour market end. Denmark needs electricians, plumbers, technicians, care workers, construction workers, and a great many people who can keep the physical and social fabric of the country functioning. It also continues to attach a great deal of cultural prestige to academic routes, even when those routes lead toward labour markets being reshaped from underneath by automation. The cultural signal that a degree is the responsible choice is colliding with a labour-market reality that increasingly rewards practical capability over abstract credentials, and the system is taking some time to register the collision.
The third is the steady rewiring of what young people use schools for. AI mental health tools are now being deployed inside school systems in the United States and elsewhere, sometimes supporting a single counsellor across three hundred students, sometimes serving as the first line for young people who would otherwise share nothing with anyone at all. A 2025 Common Sense Media survey found that seventy-two per cent of American teenagers had used AI companions, and that nearly a third had discussed serious issues with them and found those conversations at least as satisfying as friendships with other humans. Whatever you make of that, it tells you something important about what young people are quietly asking school to be: not only a place of instruction but a place of belonging, regulation, repair, and care. The teaching is in some sense the easy part.
The fourth tension is the one that arrives whenever a technology becomes cheap and abundant. AI tutors are quickly becoming adaptive enough, available enough, and inexpensive enough to make machine-led personalised instruction the rational default for stretched school systems. That makes the most personally invested form of education — being taught by an actual human who knows you, listens to you, and cares whether you grow — the thing that wealth and luck will increasingly purchase. The next inequality in education may not run along the line between those who have technology and those who do not. It may run along the line between those who are taught by people and those who are taught by software, with the social consequences of that quietly accumulating over a generation.
The fifth, and possibly the most underrated, is sustainability. According to BCG and others, fewer than one in five companies with concrete sustainability targets currently have the in-house skills to meet them, and somewhere in the order of a hundred and fifty million workers globally are estimated to need sustainability upskilling this decade. Schools and universities, by and large, are still treating this as an elective, the way they treated digital skills in the early 2000s. The catch-up is going to be painful, and the cost will not fall on the institutions that delayed it.
These tensions do not stand in a line. They pull at one another. AI both undermines the credential and makes its replacement uncertain. The mental-health pressure on schools is shaped by the same platform dynamics that are reshaping what counts as a labour market. Sustainability touches every discipline and is currently institutionally homeless in most of them. Anyone trying to address these tensions one at a time, in policy briefs or curriculum reviews, is missing how much of the difficulty is in the interaction.
What the system has quietly stopped doing
Beneath the visible tensions, there is a quieter shift worth naming. The older idea of education in much of Northern Europe was not really about training. The Danish word dannelse, and its older German cousin Bildung, describes a lifelong cultural process through which a person is shaped, matures, and develops personally, morally, and intellectually. The point was never only to know things or to do things. The point was to become a particular kind of person, capable of inhabiting a society as something more than a productive unit. Citizen as an active verb, not as a passive category.
Most contemporary education systems have not formally abandoned this idea. They have simply, over decades, stopped organising themselves around it. Performance metrics, ranking systems, employability data, completion rates, graduate earnings — none of these are bad measures, but none of them measure dannelse. The cumulative effect is that schools and universities increasingly act as sorting and credentialing machines that occasionally also produce dannelse as a side-effect, rather than as institutions whose first job is the cultural formation of a person.
This matters now because the credential side of the equation is precisely the part being eroded by AI, automation, and the loosening of the diploma-to-job pipeline. If education has been quietly running on credentialing while dannelse has been doing background work that nobody was paying attention to, the credentialing side is the side under structural pressure. What is left is the part that the system stopped describing as central, even as it is becoming central again.
Eric Hoffer's old line is doing a lot of useful work in this context: "In times of change, the learners will inherit the world, while the knowers will be beautifully equipped for a world that no longer exists." If a degree is no longer a stable passport, the question becomes which capabilities make a person able to keep learning, keep adapting, keep being a citizen in a society whose shape will not sit still. That is closer to dannelse than to most contemporary curricula.
The competitor that is not another school
There is a useful question Reed Hastings once asked about Netflix that travels surprisingly well here. He famously said the company's real competitor was not HBO or Disney but sleep, because sleep was the thing actually competing for the user's evening. Education has a similar question it has been slow to ask. Who and what is actually competing for the attention, identity formation, and sense of belonging of the people moving through the system?
The honest answer in 2026 is that the competition is everywhere, and most of it is not other schools. It is YouTube, TikTok, AI companions, Discord communities, MOOCs, capability passports, bootcamps, employer academies, fellowships, peer-learning communities, cohort-based courses, professional circles, festivals, residencies, and an entire informal ecology of learning and meaning-making that did not exist a generation ago and that is generally moving faster than the institutions. Some of this ecology is excellent. Some of it is exploitative. Almost none of it is providing the slow, structured, relational work of dannelse, because almost none of it is set up to.
For schools and universities, the competition question reframes the strategy question. If knowledge is now abundant, the value of a place of learning is not primarily that it transmits it. If credentials are weakening, the value is not primarily that it stamps you. The relatively distinctive things schools and universities can still do — slow attention, dialogue across difference, supervised disagreement, mentorship across generations, communal ritual, the deliberate cultivation of judgement — happen to be the things the informal ecology is least good at, and the things a healthy democracy keeps needing. That is closer to a strategy than most current institutional reform documents are getting.
What this asks of leaders, designers, and policymakers
If the bargain has broken, the practical implication is that the next decade of education work is going to involve a lot of people making explicit choices that the system used to be able to leave implicit. Which capabilities does this institution actually develop, and which does it merely sort for? Which parts of what we do are about credentialing, which are about transformation, and which are about belonging — and have we been honest about the weight we give each? When AI can deliver knowledge transfer cheaper and faster than we can, what exactly is the value we are charging tuition for? When sustainability is becoming a baseline rather than a specialism, where does it actually live in the curriculum, and who teaches it?
These are not abstract questions. We have run versions of them with university leadership groups, with NGOs working with young people, with public-sector clients trying to understand lifelong learning, and with foresight training participants who are themselves teachers. They are uncomfortable questions because they expose the assumptions an institution has been quietly running on, and the honest answers are rarely flattering to the current setup. But the question of what education is for becomes easier, not harder, once people are willing to admit it had stopped being clear.
A different kind of return
The conclusion that keeps emerging from this work is not a tidy prescription, and it is not the call to embrace technology that you might expect from people who spend their time looking at the future. It is something closer to a return — a return to the older question of what kind of person a society is trying to shape, and what kinds of institutions are actually equipped to do that shaping at scale. The technology is part of the picture. The labour market is part of the picture. But neither of them, on their own, will produce an answer to the question that is actually under all the others, which is what education is supposed to be for now that the postcard bargain has stopped holding.
The signals on the surface — AI in the classroom, the credentials crisis, the skilled-worker shortage, the mental-health pressure, the slow disappearance of the humanities into outsourced text — are all worth tracking, and we will keep tracking them. But the deeper story is the quiet return of an older question that the system thought it had safely answered. It had not. It had only made it possible not to ask. The institutions that work out how to ask it well, and to organise themselves around the answer, are going to look quite different a decade from now. The ones that do not are going to find themselves competing for relevance with software, on terms that were never going to favour them.
