After Scarcity: Rethinking the University in the Age of AI
This article is part of a four-part Frontiering series exploring what happens to higher education when intelligence, explanation and content generation become abundant. Rather than asking simply how universities should use AI, After Scarcity asks a more fundamental set of questions: who retains power when knowledge is commercialised, what happens to originality when everyone has access to similar intelligence, why judgement and foundational knowledge may become more important rather than less, and ultimately what a university is for when answers are no longer scarce. Across the four essays, the argument moves from power → originality → judgement → purpose.
4. The University in an Age of Abundant Intelligence - Purpose
What becomes valuable when answers are no longer scarce?
For most of their history, universities have existed in a world of scarce intelligence.
Books were scarce.
Experts were scarce.
Information was difficult to find.
Sophisticated analysis required years of education.
If you wanted access to specialised knowledge, you went somewhere that possessed it.
The university.
Its libraries contained knowledge you could not access at home. Its academics knew things you did not know. Its laboratories contained equipment you could not afford. Its seminars placed you among people capable of challenging your thinking.
And its degree provided evidence that you had spent several years acquiring knowledge and learning how to use it.
Artificial intelligence changes some of those assumptions rather dramatically.
A student can now carry something in their pocket capable of explaining quantum mechanics, critiquing a marketing strategy, writing Python, translating Mandarin, analysing a balance sheet, discussing Kant, creating a research proposal and tutoring them patiently at two o'clock in the morning.
None of this means the university has become obsolete.
But it does mean we should probably stop pretending that nothing fundamental has changed.
The question confronting higher education is no longer merely:
How should universities use AI?
It is a much bigger question:
What is a university for when intelligence itself becomes abundant?
We may be solving yesterday's scarcity problem
Much of higher education still operates around a familiar model.
An expert knows something.
Students do not.
The expert organises that knowledge into a course.
Students attend lectures, read material, participate in activities and demonstrate that they have acquired some of it.
At the end, the university certifies the result.
Of course, good universities have always done considerably more than transmit information.
But knowledge transmission remains deeply embedded in how universities are organised:
lectures;
textbooks;
learning-management systems;
content modules;
tutorials;
examinations;
essays;
reports; and
degrees constructed largely as sequences of subjects.
All of these developed in an environment where access to expert knowledge was relatively expensive.
AI dramatically lowers that price.
And universities are adopting it quickly.
In a 2026 EDUCAUSE survey of 1,960 higher-education professionals, 92 per cent of respondents said their institution had some form of work-related AI strategy. Yet only 13 per cent said their institution was measuring the return on its AI investment (Robert, 2026).
UNESCO similarly found that 19 per cent of surveyed higher-education institutions already had formal AI policies, while another 42 per cent were developing them. Importantly, some universities were moving beyond rules about AI use towards redesigning assessment and introducing AI literacy into curricula (UNESCO, 2025).
So adoption is clearly occurring.
But adoption is not transformation.
Giving everybody an AI licence does not answer the more important question of what becomes valuable once everybody has one.
When something becomes abundant, value moves elsewhere
Economics teaches us to pay attention to scarcity.
For centuries, access to information and sophisticated intellectual assistance was scarce.
Universities accumulated enormous value around that scarcity.
AI does not eliminate the need for knowledge.
Quite the opposite.
As Manu Kapur (2026) argues, plausible answers can now be produced so quickly that shallow domain knowledge may become more dangerous. The important capability is increasingly not obtaining an answer but possessing enough knowledge to determine whether that answer is any good.
But AI does change where scarcity sits.
Consider what is becoming inexpensive.
Explanation
Ask a capable AI to explain almost any established concept at almost any level of complexity.
Summarisation
Hours of material can be condensed in seconds.
First drafts
Essays, reports, emails, computer code, presentations and proposals can be produced almost instantly.
Translation
Language is becoming much less of a barrier to accessing knowledge.
Basic analysis
Large volumes of structured and unstructured information can increasingly be examined conversationally.
Personalised assistance
One-to-one explanation, practice, brainstorming and feedback can increasingly be available whenever someone wants it.
None of these capabilities is perfect.
But perfection is not required to change the economics.
When something previously scarce becomes abundant, the scarce things around it become relatively more valuable.
And I suspect that is where the future university needs to look.
Judgement is scarce
AI can generate twenty strategies.
Someone still needs to decide which one makes sense.
It can identify ten possible diagnoses.
Someone still needs to decide what to do.
It can construct persuasive arguments for opposing positions.
Someone still has to determine which argument is supported by the evidence, which values matter and what consequences should follow.
Judgement is not simply another output.
It is the capacity to act under uncertainty when several apparently reasonable possibilities exist.
That becomes more valuable, not less, when generating possibilities becomes cheap.
Perhaps universities should therefore spend less time asking:
Can the student produce the answer?
and considerably more time asking:
Can the student decide what to do with the answers available to them?
Originality is scarce
There is another potential consequence of abundant machine intelligence.
If everyone uses increasingly similar systems trained upon much of the same accumulated human knowledge, our capacity to generate plausible ideas may increase while our outputs become more alike.
Recent research into US research funding provides an intriguing early warning.
Qian et al. (2026) examined confidential NIH and NSF research proposals alongside publicly released awards. Greater estimated LLM involvement was consistently associated with lower semantic distinctiveness — projects appeared closer to work that agencies had already funded. At NIH, greater LLM involvement was also associated with higher proposal success and more subsequent publications, although the additional output was concentrated in papers that were not among the most highly cited.
That does not mean AI causes unoriginal research.
But it does raise a fascinating possibility.
AI may help each of us become individually more productive while collectively nudging us towards what has already worked.
If so, one of the university's most valuable functions becomes protecting intellectual variance.
The unusual hypothesis.
The interdisciplinary connection.
The awkward question.
The idea that initially looks slightly ridiculous.
The researcher prepared to say:
Everyone seems to be looking over there. What happens if we look over here?
When intelligence becomes abundant, originality becomes scarce.
Attention is scarce
Universities face another scarcity that has little to do with artificial intelligence itself.
Attention.
Students already inhabit an environment engineered to interrupt it.
Notifications.
Short-form video.
Messages.
Infinite scrolling.
Multiple screens.
AI introduces another paradox.
It enables us to process enormous quantities of information while making it increasingly possible never to engage deeply with any of it.
We can summarise the article instead of reading it.
Summarise the summary.
Generate the questions.
Generate the answers.
Generate the reflection about the answers.
At some point, we should probably ask who is actually learning.
A university may therefore increasingly need to provide something surprisingly old-fashioned:
protected spaces in which people pay sustained attention to difficult things.
A seminar in which twelve people spend an hour struggling seriously with one complicated question may become more educationally valuable precisely because an AI can generate an immediate answer.
The struggle may be the point.
Trust is scarce
There is another consequence of cheap intelligence.
Convincing text is cheap.
Images are cheap.
Audio is cheap.
Video is becoming cheap.
Synthetic data can be generated cheaply.
Plausibility, therefore, becomes a weaker signal of truth.
The important questions increasingly become:
Who produced this?
What evidence supports it?
What has been omitted?
What assumptions are hidden?
Can I trust the source?
Could this have been fabricated?
Universities have always been institutions of knowledge.
Increasingly, perhaps they need to think of themselves as institutions of epistemic trust.
Places where people learn not only what we know, but how reliable knowledge comes to be known.
That requires:
evidence;
methods;
replication;
peer challenge;
citation;
argument;
transparency; and
the willingness to change one's mind when the evidence changes.
These are not quaint academic conventions.
In a world filled with synthetic knowledge, they become societal infrastructure.
Interestingly, higher-education professionals themselves are already identifying these risks. In Robert's (2026) EDUCAUSE study, misinformation, non-consensual data use and loss of skills requiring independent thought were among the most frequently identified urgent concerns surrounding AI. Respondents also raised concentration of wealth and power, loss of creativity, and surface-level work.
Human relationships are scarce
AI can tutor.
It can coach.
It can provide feedback.
It can simulate a remarkably patient conversation.
But education has never simply been an information exchange.
Students develop partly through encounters with other people.
The academic who sees potential in them.
The classmate who completely disagrees.
The team member who fails to deliver.
The supervisor who asks the question they hoped nobody would ask.
The client whose problem refuses to fit neatly into the framework.
The colleague from another discipline who sees the problem completely differently.
These encounters create friction.
And friction is often where learning occurs.
One of the great mistakes universities could make would therefore be responding to AI by making education even more digital, individualised and frictionless.
The more machine intelligence becomes available, the more distinctive meaningful human interaction may become.
Real problems are scarce
AI performs particularly well when the problem has already been defined.
The world is much less cooperative.
Organisations rarely arrive saying:
Here is a perfectly specified problem, complete with all the relevant variables, stakeholder priorities and a reliable dataset.
They arrive with:
incomplete information;
conflicting stakeholders;
political constraints;
unclear objectives;
imperfect data;
competing values; and
people who disagree about what the problem actually is.
This suggests another shift.
Perhaps the fundamental educational unit should increasingly become not the content module, but the authentic problem.
Give students a real organisation.
A real community.
A real dataset.
A real ethical dilemma.
A real client.
A real decision.
And let them use AI.
Then assess what they do with it.
That looks considerably more like the world graduates are entering anyway.
Perhaps the university moves from transmission to formation
I think this is the larger change.
A university built around scarce information naturally emphasises transmitting knowledge.
A university operating amid abundant intelligence needs to emphasise forming capable people.
People with:
judgement;
disciplinary knowledge;
curiosity;
intellectual independence;
ethical reasoning;
creativity;
collaborative capability;
resilience;
evidence literacy;
communication skills;
the ability to frame problems; and
the confidence to disagree intelligently with both humans and machines.
Knowledge remains central.
But its purpose changes.
We learn history not simply because historical facts are difficult to retrieve, but because history gives us patterns through which to understand the present.
We learn statistics not simply because calculations are difficult, but because statistics allows us to recognise when numbers are misleading us.
We learn philosophy not because an AI cannot explain utilitarianism, but because societies still require humans capable of reasoning about what ought to be done.
We learn management not simply to memorise frameworks, but to understand when a framework helps, when it doesn't, and what to do when people refuse to behave like the framework predicts.
We learn a profession not simply to know its accumulated answers.
We learn it to acquire the judgement necessary when established answers no longer quite fit.
Assessment has to change with it
If the finished artefact can increasingly be generated cheaply, universities cannot continue assigning most of the educational value to the artefact itself.
The essay still has value.
The report still has value.
The presentation still has value.
But increasingly we need to examine the intelligence surrounding its creation.
What question did the student ask?
What assumptions did they identify?
What evidence did they trust?
What did they reject?
Where did AI influence the work?
Where was the AI wrong?
What alternative explanations did they consider?
Why did they ultimately decide what they decided?
Assessment shifts from:
Show me what you produced.
towards:
Show me how you exercised judgement while producing it.
That is harder to assess.
It is also much closer to professional life.
The campus may become more important, not less
There was a moment during the rise of online learning when it seemed possible that digital content would dramatically reduce the importance of physical universities.
AI might initially appear to strengthen that argument.
If everyone possesses a personalised tutor, why attend campus?
I suspect the opposite may eventually occur.
If information and explanation become abundant digitally, the things that physical universities can uniquely provide become relatively more valuable:
laboratories;
studios;
clinics;
workshops;
performances;
debate;
mentoring;
communities;
friendships;
professional networks;
cultural experiences;
interdisciplinary encounters; and
access to people doing genuinely new things.
The mediocre lecture becomes harder to justify.
The excellent university becomes easier to justify.
Because its value is no longer principally the content contained within the lecture.
It is the environment in which people become something they could not easily become alone.
Universities must also decide what they themselves need to know
There is an institutional version of this argument.
Universities increasingly depend upon AI systems developed by a handful of technology companies.
That relationship can deliver tremendous value.
But if universities outsource their AI capability entirely, they risk becoming intelligent organisations that understand surprisingly little about the intelligence upon which they depend.
Universities do not need to build everything themselves.
But they probably need meaningful internal capability around:
institutional data;
AI governance;
model evaluation;
academic intellectual property;
open models;
procurement;
research infrastructure;
student privacy; and
understanding how AI influences teaching, research and institutional decision-making.
Robert's (2026) findings are interesting here too: while 92 per cent of respondents said their institutions had AI strategies, only 54 per cent were aware of policies and guidelines governing their own AI use, and 56 per cent reported using AI tools not supplied by their institutions.
Adoption is moving more quickly than institutional understanding.
That deserves attention.
Because abundant intelligence does not automatically mean distributed power.
And perhaps credentials have to change
There is one final question.
What exactly should a university degree certify?
Historically, a degree has operated partly as evidence that somebody knows particular things and can produce particular forms of intellectual work.
That remains important.
But if AI increasingly participates in that work, the signal needs to become more precise.
Perhaps a graduate credential should increasingly communicate something like:
This person possesses sufficient foundational knowledge to work intelligently in this field.
This person can use powerful AI systems without surrendering judgement to them.
This person can distinguish evidence from plausibility.
This person can collaborate with other people.
This person can frame difficult problems.
This person can make and defend decisions.
This person understands the ethical consequences of those decisions.
And this person can recognise when they do not know enough.
That would be an extremely valuable qualification.
The opportunity is bigger than defending the old university
Much of the discussion about AI in higher education remains defensive.
How do we prevent cheating?
How do we detect AI?
Which tools should students be allowed to use?
What should our policy say?
These are legitimate questions.
But they are small questions compared with the opportunity.
Artificial intelligence is forcing universities to confront something they should probably confront periodically anyway:
Purpose: What is genuinely valuable about what we do?
If the answer is merely that universities possess information and experts capable of explaining it, AI presents a serious challenge.
But I don't think that is the answer.
The university at its best is not a warehouse of answers.
It is a place where people learn how knowledge is created.
Where ideas collide.
Where assumptions are challenged.
Where novices encounter experts.
Where disciplines preserve deep bodies of knowledge while occasionally overturning them.
Where people learn to think with other people.
Where society creates protected space for questions whose answers may not be commercially useful tomorrow.
And where people develop the capacity to make consequential decisions in a complicated world.
AI does not make those functions less important.
It may make them considerably more important.
So perhaps the university in an age of abundant intelligence should stop competing with AI at the things AI makes cheap.
And concentrate instead upon the things that remain scarce.
Judgement.
Originality.
Attention.
Trust.
Human connection.
Experience.
Curiosity.
Wisdom.
And perhaps most importantly:
knowing which questions are worth asking in the first place.
References
Kapur, M. (2026, July 28). AI makes foundational knowledge more important than ever. Times Higher Education.
Qian, Y., Wen, Z., Furnas, A. C., Bai, Y., Shao, E., & Wang, D. (2026). The rise of large language models and the direction and impact of US federal research funding. Proceedings of the National Academy of Sciences, 123(33). https://doi.org/10.1073/pnas.2601439123.
Robert, J. (2026, January). The impact of AI on work in higher education. EDUCAUSE.
UNESCO. (2025, September 2). UNESCO survey: Two-thirds of higher education institutions have or are developing guidance on AI use.