Universities are entering a pivotal moment with generative Artificial Intelligence, compared to the arrival of the pocket calculator in classrooms in the 1970s. The core lesson from that earlier shift was that faster machines did not remove the need for human thought. They increased the value of judgment, number sense, and the ability to recognize when an answer was wrong. A similar transition is now unfolding in higher education, not as the end of universities, but as a redefinition of what learning is for.
Generative Artificial Intelligence has moved into universities with unusual speed. ChatGPT reached 100 million users in just two months. By 2025, more than a billion people were using Artificial Intelligence tools globally. Higher education has become a major testing ground because universities center on tasks that large language models handle well, including reading, writing, summarising, synthesising, and generating arguments. That creates both opportunity and disruption. Artificial Intelligence opens the possibility of scalable personalised learning that universities have long been unable to deliver broadly.
One of the clearest opportunities is tutoring. For decades, education researchers have understood that one-to-one tutoring can significantly improve student outcomes. Benjamin Bloom described this as the “2 sigma problem”: students with personalised tutoring often perform two standard deviations better than those in conventional classrooms. The historical barrier was cost. Universities can now imagine giving every student a 24/7 Artificial Intelligence tutor capable of scaffolding learning, adapting explanations, generating practice activities, and providing immediate feedback at scale.
At the same time, generative Artificial Intelligence puts pressure on long-standing assumptions about assessment, authorship, and intellectual development. The university essay has traditionally served as evidence of thinking, but students can now generate polished assignments quickly without necessarily understanding the material. That raises a more fundamental question than academic misconduct alone: how universities can tell whether learning has actually happened. Treating Artificial Intelligence mainly as something to detect or prohibit risks missing the deeper issue, which is that the technology exposes existing weaknesses in assessment design.
The response proposed is to shift attention from finished products to the learning process itself. Universities are being pushed to make thinking more visible, value iteration and reflection, redesign assessments around judgment and application, combine supervised and unsupervised assessment, and teach students to use Artificial Intelligence critically rather than pretend it does not exist. The larger challenge is cultural as much as technical. Universities are urged to avoid both utopian claims that Artificial Intelligence will solve education and dystopian fears that it will destroy it, and instead focus on combining technological capability with human judgment, motivation, trust, mentorship, belonging, and social interaction.
