Who owns the time AI saves? New politics of academic labour

For most of the university’s history, intelligence was costly.

To make expertise accessible to students, universities had to employ people who had spent years acquiring it. To evaluate an argument, supervise a dissertation, design a curriculum, or explain a difficult concept, institutions needed academics whose knowledge could not be easily separated from the people who possessed it.

That economic assumption shaped the university as much as any philosophy of education did.

Artificial intelligence is beginning to disturb it.

The immediate debate has centred on familiar questions. Will students use AI to write essays? How should assessment change? Can professors use generative systems to prepare lectures? What constitutes plagiarism when writing is increasingly produced through human-machine interaction?

The deeper question is what happens when universities discover that some forms of academic cognition can be reproduced, scaled and delivered without requiring a human each time.

Who redesigns academic labour in the age of AI?

That question will increasingly arise in faculty contracts, workload negotiations, intellectual-property disputes, promotion systems and institutional governance.

The coming argument about AI in universities may therefore be less about machines replacing professors than about something subtler: the gradual decomposition of the professor into functions that institutions can reorganise, redistribute and perhaps eventually purchase separately.

AI doesn’t need to replace professors

The phrase ‘AI will replace professors’ is likely misleading.

Professors are unlikely to suddenly disappear from universities. Academic institutions depend on accreditation, disciplinary authority, mentorship, research supervision, institutional governance and human accountability in ways that cannot be easily transferred to a chatbot.

But replacement need not occur at the job level. It can occur at the task level.

A contemporary academic position is essentially a set of responsibilities. A professor prepares lectures, designs courses, answers questions, evaluates assignments, advises students, provides feedback, writes reports, conducts research, serves on committees, supervises projects and participates in academic governance.

AI does not need to perform all of these tasks.

If course preparation becomes substantially faster, universities may increase expected teaching loads.

If AI tutors handle routine questions, universities may need fewer teaching assistants. If assessment becomes partially automated, a single faculty member may oversee significantly larger cohorts.

If introductory courses can be generated centrally and delivered across multiple campuses, institutions may need fewer faculty members designing similar courses independently.

The unit of automation is not necessarily the occupation. It is the task. But when enough tasks are automated, the occupation itself is redesigned.

The first danger may be productivity, not unemployment. AI is typically introduced as an assistive tool: preparing lectures, speeding grading, generating materials, answering routine questions and reducing administrative burdens.

But technologies that increase worker productivity have a historical problem: the time saved does not necessarily return to the worker.

Suppose an AI system allows a professor to complete in eight hours what once took 10.

Who owns the two hours saved?

That apparently simple question contains much of the future of academic AI politics.

Whether the time saved returns to faculty for research, mentoring, or intellectual development – or is used to expand class sizes, add courses, or reduce positions – depends on institutional power, not technology.

Without negotiated answers, AI may create one of the paradoxes of modern academic life. A technology introduced to reduce workload could become the very mechanism that expands workloads.

What appears to be technological progress may therefore become labour intensification.

This pattern is not unique to universities, but higher education has an additional complication. Academic labour is already unusually elastic. Research extends into evenings. Student communication extends into weekends. Administrative expectations accumulate gradually. Faculty rarely encounter a clear boundary marking the end of the workday.

AI could become another instrument that further stretches that elasticity.

Professors are sources of training data

Universities hold extraordinary reservoirs of faculty-generated intellectual material: recorded lectures, course notes, presentations, assessment rubrics, feedback, discussion board responses, advising records, research seminars, online courses and countless explanations preserved in email.

Twenty years of teaching can leave an enormous digital trace of how a particular academic thinks.

Until recently, those materials were mainly archives. AI can transform them into capabilities.

Imagine a professor who has taught economics for 25 years. The university has hundreds of recorded lectures, thousands of student interactions, examinations, explanations, examples, comments and course materials.

An AI system trained on those materials could potentially answer questions in recognisably similar ways, generate examples based on the professor's teaching patterns, or support students using a body of knowledge accumulated through decades of academic labour.

What exactly has the university acquired?

Course materials? Intellectual property? Or a partial computational representation of the professor? This is not a trivial distinction.

Traditional intellectual property agreements usually ask who owns a book, a lecture recording, or a piece of software.

AI introduces a different possibility.

The intellectual product does not simply remain a product.

It can become a machine for producing additional products.

A lecture becomes a training asset.

Feedback becomes a feedback system.

A teaching archive becomes an instructional model.

The economic value of faculty knowledge therefore shifts.

Academics may be the first generation of university workers whose accumulated intellectual labour can be transformed into systems that reduce the institution's future dependence on that labour.

Faculty members could, in effect, help train the infrastructure that makes parts of their own work optional. That possibility should fundamentally reshape universities' approach to consent, intellectual property and compensation.

The relevant question is no longer merely: ‘Who owns my lecture?’ It is: ‘Who owns what can be built from my lecture?’

Most existing faculty contracts were not designed to address that question. The professor could become a supervisor of synthetic academic labour. Perhaps AI does not replace the professor. Perhaps it multiplies the professor.

A single academic might eventually oversee a constellation of AI tutors, automated feedback systems, course agents, advising tools and research assistants, all operating under the professor’s authority.

One professor could therefore reach far more students than is possible today. The professor becomes less of a direct educator and more of a manager of synthetic academic labour. That could create a new hierarchy within the university.

Highly valued academics might design intellectual systems used by thousands of students, while other academics become supervisors of standardised AI-mediated courses.

Prestigious institutions could distribute the intellectual output of prominent faculty globally, while less wealthy universities purchase access rather than maintain equivalent academic capacity locally.

AI can democratise access to knowledge while centralising control over its production.

Most discussions position AI below the professor: an assistant, tutor, writing tool, or marking aid. But AI can also move upward.

Universities already collect vast amounts of information about academic employees: student evaluations, publication data, citation metrics, grant income, completion rates, course enrolments, online activity and other performance indicators.

AI makes it possible to integrate these data into increasingly sophisticated systems for evaluating faculty productivity.

What happens when AI does not merely perform academic work but begins judging those who do?

A university might claim that an algorithm merely ‘supports’ decisions about promotion, hiring, workload, or performance. Yet decision-support systems have a curious tendency.

When they become sufficiently sophisticated, people start deferring to them. The algorithm becomes advice. Advice becomes recommendation. Recommendation becomes default. And eventually a human administrator signs a decision that was substantially prepared elsewhere.

Universities should be especially cautious here. Academic life depends on forms of value that are difficult to quantify.

The professor who transforms a struggling student's intellectual trajectory may not produce any meaningful productivity indicator. The scholar who spends 10 years developing an unconventional idea may seem inefficient until the idea becomes important. The teacher who rigorously challenges students may receive lower evaluations than one who makes few demands.

Academic work contains forms of delayed value, relational value and intellectual risk that computational management systems may systematically undervalue. There is therefore a boundary worth defending:

AI may help us see academic work. It should not be allowed to determine what academic work is worth. Otherwise, universities risk allowing their measurement systems to become their philosophy.

Weakest academic workers may experience the future first

Focusing solely on tenured professors would be a mistake. The academic labour ecosystem also includes adjunct instructors, teaching assistants, tutors, advisers, instructional designers, language support staff, research assistants, librarians and administrative professionals.

These occupations may face substitution pressures before professors do. Organisations rarely begin automation by replacing their most powerful employees. They automate around them.

A university may retain the professor while eliminating some of the people who once surrounded the professor. The visible classroom remains human.

The infrastructure beneath it becomes increasingly synthetic. AI tutoring reduces demand for tutors. Automated advising reduces entry-level advising positions. Automated assessment changes teaching assistants’ workloads.

Generative course systems reduce some forms of instructional design work. Research agents alter the work research assistants perform. Replacement is rarely evenly distributed. It follows institutional power.

The people least able to negotiate the terms of automation are often the most exposed to it. Therefore, faculty labour negotiations cannot focus solely on protecting faculty. They must ask what kind of human university faculty themselves are willing to preserve.

It is no longer human vs machine

At this point, it is tempting to defend a simple principle: education should remain human. But that argument is too easy.

There are many things AI may eventually do extremely well: explain difficult material patiently, provide feedback at any hour, translate across languages, give students private space to practise basic questions and help instructors identify misconceptions across large cohorts.

Refusing these capacities merely because they are artificial would make little sense. The important distinction is not between human and machine but between capability and responsibility.

An AI system may become capable of providing an excellent explanation. But can it assume responsibility for a student's education?

It may recommend that a student is underperforming. But can it be accountable for the consequences of that judgement?

It may generate a curriculum, but can it defend the intellectual values embedded within that curriculum before a faculty senate, a profession, or a society?

Universities should not confuse the ability to perform an intellectual act with the authority to assume responsibility for it. That distinction may ultimately define the line between academic augmentation and academic substitution.

Faculty should negotiate authority

Future collective bargaining over AI should therefore move beyond clauses stating that professors cannot be replaced and instead negotiate the authority over automation itself.

Who decides which educational tasks can be automated?

Who determines whether an AI system is pedagogically appropriate?

Who controls faculty-generated data?

Who benefits from productivity gains?

Who may create AI models using course materials?

Can automated systems evaluate academic work?

When must students be informed that they are interacting with AI?

Who remains accountable when an AI-supported academic decision causes harm?

These are not merely employment questions. They are constitutional questions about the university. For decades, academic governance has rested on the principle that professors possess professional authority over teaching, curriculum and disciplinary judgement.

If AI systems become embedded in those domains, control over AI becomes inseparable from academic governance. A faculty contract that negotiates salary but ignores computational authority may soon address only part of the employment relationship.

The need for productivity dividend conversations

If AI genuinely makes academic work more efficient, who should receive the benefit?

AI-generated efficiencies could instead fund smaller classes, additional research time for faculty, lighter advising loads and more individual mentoring, while allowing universities to reinvest some of the savings in precarious academic workers rather than eliminating their positions.

AI could allow academics to spend less time producing routine content and more time on what universities claim to value: mentoring, inquiry, intellectual experimentation and community.

In other words, AI could finance a more human university. But that outcome will not emerge automatically. Technologies do not distribute productivity gains. Institutions do.

The more important question is not simply how many professor positions AI might eliminate, but which parts of professorial work actually deserve to survive.

Do academics really need to spend hours formatting routine documents?

Should professors repeatedly answer the same administrative questions?

Is manually producing hundreds of similar comments the highest use of academic expertise? Probably not.

AI gives universities an opportunity to remove work that was never intellectually central to the professoriate. But there is a danger. Institutions may automate the wrong thing.

They may preserve bureaucracy while automating intellectual relationships. They may keep administrative complexity while eliminating small seminars. That would be technologically efficient and educationally absurd.

The goal, therefore, should not be to preserve every existing faculty task.

It should preserve what makes academic labour academically meaningful: judgement, mentorship, intellectual risk, disciplinary memory, ethical responsibility and the ability to recognise a student as more than a pattern in data.

And perhaps most importantly, the capacity to say: ‘This conclusion may be efficient, measurable and computationally defensible – and still be wrong.’

The university negotiates with intelligence

The first generation of university technology digitised information. The next generation connected people. AI is different. It participates in cognition.

When universities automate payroll, no one asks whether the institution still needs human judgement.

When universities automate parts of teaching, assessment, advising, research and intellectual production, the question becomes unavoidable. We are no longer negotiating solely over adopting a tool. We are negotiating the boundary between institutional intelligence and human responsibility.

The future academic labour agreement may therefore have to do something no previous faculty contract was asked to do. It may have to specify which forms of cognition belong to machines, which belong to professionals, which may be shared between them and where final authority must remain human.

The decisive question will then be not whether AI is intelligent enough to participate in the university. It already is. The real question is whether universities will be wise enough to decide which human capacities become more valuable when intelligence itself is no longer scarce.

That is the negotiation: higher education should begin before the technology begins negotiating it for us.

James Yoonil Auh is a former vice-president of the National Labor College in the US and vice-president of Kyung Hee Cyber University in Seoul, Korea. He currently teaches at Judson University in Illinois. His work examines how technological and social change reshape institutions, authority and human agency.

This article is a commentary. Commentary articles are the opinions of the author and do not necessarily reflect the views of University World News.