Artificial intelligence is reshaping education at warp speed. From personalised learning pathways to instant feedback and administrative support, AI tools already reside inside classrooms across the world. From the U.S. to India, adoption figures reflect a clear reality: the question is no longer whether AI belongs in education (it already does) but whether we are using large language models to optimise learning.

Benefits and risks

AI’s benefits are unequivocal, and a flattened world of learning is the most obvious. A student in a remote town — or in a school with absent or indifferent teachers — can now receive instant clarification on complex topics that once depended on geography, privilege, or access to specialised tutors.

AI also enables personalisation at a scale that traditional classrooms struggle to achieve. It can adapt pace, difficulty, and examples to individual learners — something a single teacher managing 50 or 60 students cannot do effectively. Students often ask bolder questions when not performing publicly; used thoughtfully, AI can encourage curiosity rather than suppress it.

Yet underneath these gains lurks a genuine risk. The danger is not that AI provides answers. Textbooks have always done that. The risk is that AI makes it effortless to obtain answers without engaging in the cognitive effort that produces understanding. Learning is not the accumulation or regurgitation of facts; it is the building of connections. It involves asking the right questions, testing assumptions, and exploring implications. Simply put, AI lowers the cost of disengagement.

A second concern involves verification and critical judgement. Learning traditionally exposes students to multiple voices: teachers, texts, peers, conflicting sources. This friction helps students evaluate credibility and uncertainty. AI systems, by contrast, often present information with smooth confidence, rarely signalling doubt or debate unless prompted. When students increasingly rely on AI for research, career guidance, or decision-making, the question is not simply accuracy; it is also about cross-checking and contextualising.

Warning signs already visible

Early warning signs are visible. Academic communities have reported a surge of superficially competent output — work produced at speeds incompatible with genuine scholarship. The issue is not assistance; it is substitution. When tools designed to support thinking begin to replace it, education loses its core purpose. Yet the problem is not AI itself, but over-reliance on AI as a shortcut rather than a scaffold.

This is where human teachers become irreplaceable. India faces a significant shortage of educators, and it becomes convenient to frame AI as a numerical solution. But teaching is not a volume problem. It is a relational one. Good teachers do far more than transmit content. They notice confusion students cannot articulate. They recognise when effort falters because of fear, fatigue, or circumstances outside the classroom. Good teachers model intellectual honesty, emphasise ethical reasoning, and foster an environment where it is okay to admit uncertainty. Good teachers mentor and shape not just what students know, but how they think — and who they become. That influence cannot be automated because mentorship is inherently human.

The importance of teachers

AI’s greatest influence in education might therefore be visible in teachers themselves. A teacher who understands when to rely on AI and when to resist it models discernment every day; over time, the multiplier effect is profound.

But discernment should not become a mere set of rules. ‘Using AI responsibly’ looks different across contexts. Brainstorming ideas with AI is different from outsourcing authorship. Drafting assessments with AI support is different from accepting outputs uncritically. Judgement is most effective when demonstrated, practiced, and discussed — not merely imposed.

This places responsibility on individuals and institutions alike. We must design learning environments where thinking is harder to outsource than copying, where process matters as much as output, and where integrity is reinforced through culture, not surveillance.

Towards AI pedagogy

Teacher development becomes pivotal. Training must move beyond basic AI literacy towards AI pedagogy: how to design assignments that demand reasoning, how to integrate AI without diluting effort, and how to speak honestly with students about appropriate use. Assessment must evolve as well. If AI can complete a task effortlessly, it is worth asking whether the task measures what truly matters.

Learning is not meant to be frictionless. Difficulty, when supported, is formative; stumbling blocks are opportunities to fall and rise. Not every inefficiency needs to be optimised away.

The debate around AI in education often collapses into extremes: uncritical adoption or outright rejection. Neither serves students well. AI is best understood as an amplifier. It strengthens thoughtful pedagogy and exposes weak pedagogy.

The task ahead is clear. Invest in teachers — not just in numbers, but in capability and stature. Address class sizes so mentorship becomes possible. Teach AI literacy as a habit of questioning, not just tool usage.

AI will remain part of education. What is not guaranteed is whether it deepens understanding or thins it. That outcome will depend on the choices we make as institutions and educators. The best teachers have always shown students where to look, not what to see. In an age of intelligent machines, that human task has never mattered more.

(The author is Chairman, Hero Enterprise; and Chancellor, BML Munjal University. Views are personal.)