AI in universities: the case in favour

This article is part of our Prof v. Prof summer debate series, where academics take opposing viewpoints on hot topics in higher education. Read all the articles from the series here. 

Universities have long organized education around a basic constraint: the scarcity of expert time. To maximally leverage professors, students are taught in large groups using common lectures, readings and assignments. This model achieves scale but sacrifices the individual attention that produces the best learning. 

Education scholar Benjamin Bloom framed the search for a better method as the “2 Sigma problem,” in reference to studies showing that the average student receiving mastery-focused one-to-one tutoring outperformed 98 per cent of students in a conventional class. Yet he concluded that such tutoring was “too costly for most societies to bear on a large scale.” 

For decades, that constraint remained binding. Digital technologies made content easier to store, distribute and access through learning platforms, online videos and digital libraries. But they did little to alter the core model of expert-led, standardized instruction because teaching and feedback still depended on instructors’ tacit expertise and scarce time. 

AI changes that equation. Traditional algorithms could automate only what humans could codify. Machine learning, however, can capture tacit expertise by observing in data what experts do in different contexts and which actions produce the best outcomes. Given enough examples of how the best instructors teach and give feedback, machines can capture much of that expertise and make it available at scale. For the first time, one-to-one tutoring for every student is possible. 

In my own work, I describe how general-purpose technologies are adopted in three stages: Replace, Reimagine and Recombine. Universities remain largely at Replace, using AI to draft course materials, create quiz questions or write memos. These are useful gains, but they leave the basic model intact. The larger opportunity is to reimagine teaching around AI. 

Future learning will be personalised, mastery-based and adaptive. Content and examples will be customized for each student. Students will move at different speeds, mastering one topic before advancing to the next. Pedagogy will adapt in real time to students’ prior knowledge, misconceptions and progress. An AI tutor can identify gaps, vary explanations, generate relevant examples, provide repeated practice and adjust the level of difficulty. It can provide high-quality guidance outside office hours and far more feedback than even the most dedicated professor. 

This will shift the professor’s role from broadcaster of content to designer of learning environments. With AI handling routine explanation and practice, faculty can focus on constructing the intellectual journey of a course: posing difficult questions, leading debate, connecting ideas, mentoring students and upholding disciplinary standards. Technology can extend the professor’s reach while preserving human attention for the parts of education that depend most on experience, judgment and connection. 

Read also: AI in universities: the case against

Early evidence points to the benefits of reimagining education. A 2025 randomised controlled trial in a Harvard undergraduate physics course found that students learning through an AI tutor learned more in less time than those in an active-learning class. They also reported greater engagement and motivation. 

The AI revolution will transform how work is done and what skills are most valuable. Universities must therefore change not only how they teach, but also what they teach. Every student must become AI literate. To better leverage AI, they must understand its capabilities and learn how to formulate a problem, direct the system, check its claims, refine its work and take responsibility for the final product. They must also understand its limitations and challenges: alignment, bias, lack of explainability and the danger of systems that sound authoritative even while being wrong. 

However, stand-alone AI courses, detached from the core curriculum, are not enough. The technology must be integrated within disciplines because its opportunities and risks differ across fields. Economists can use AI to gather and analyse data. Engineers can use it to develop and evaluate designs. Health professionals should learn to work with AI-assisted diagnosis without surrendering clinical judgment. The goal is not simply to produce competent users of a tool, but competent professionals in a world where intelligent tools are embedded in practice. 

Disciplinary knowledge will remain important. Students who do not understand a field cannot pose the right questions to an AI system or recognize when its answers are shallow, misleading or wrong. Universities must combine deep disciplinary foundations with the ability to direct, question and improve the output of AI. 

Curricula must also respond to what will be a profound shift in the value of skills. AI is making competent writing, translating, coding, summarization, editing and background research increasingly abundant. These skills will no longer command high wages. Greater value will accrue to skills that complement AI: problem framing, critical thinking, judgment, leadership and entrepreneurship.  

Paradoxically, integrating AI into curricula and pedagogy should make the university more human, not less. As content, explanation and feedback become abundant, discussion, debate, mentorship, collaboration and shared experience become more valuable. Residences, clubs, athletics, design teams, student government, laboratories, entrepreneurship programs and co-op placements will be central to the AI university because they develop the judgment, leadership, initiative and interpersonal skills whose value will rise in an AI-rich world.  

For centuries, the university has been shaped by the scarcity of expert time. AI relaxes that constraint for the first time. The task now is not simply to add AI to the existing model, but to reimagine curricula and teaching around what it makes possible. Done well, this transformation will not diminish the human dimensions of education. It will create more time and space for them.