When AI can teach, what is the teacher for?

This article is authored by Rahul Ramachandran, director, Partnerships and Initiatives, KRUU.

Indian classrooms are already incorporating AI. According to the Bharat Survey for EdTech 2025, which involved 2,500 teachers and 12,500 households in ten states, 35% of kids who use educational technology already use generative AI tools. 51% of educators who use educational technology said they use generative AI for instruction and school-related tasks.

AI, teachers and learning

The majority of what we have historically referred to as teaching—explaining, drilling, and correcting—will soon be performed by AI. This does not diminish the importance of the teacher. By assisting students in framing real-world problems, gathering information that no machine can provide, and determining which answers are reliable, it takes the teacher to the real-world learning environment. Schools will prosper if they reorganise around this. Schools that merely incorporate AI into the current model will speed it up rather than improve it.

In practice, that assertion is more evident than in theory.

In a recent project we conducted, students asked if their neighbourhood could be transformed into a 15-minute city, where everyday necessities could be met with a quick stroll or bicycle ride, for thirty days.

AI would have been a great collaborator for brainstorming. In a matter of seconds, it could describe the idea, provide an overview of how Paris or Barcelona used it, and recommend a dozen designs. It couldn't tell a student how far the closest clinic is from her building, where the path ends, or how many km her family travels each week. That had to be collected on foot using personal footprint calculations, travel journals, and hand-drawn neighbourhood maps.

It was necessary to do that work before consulting AI. A student cannot ask a pertinent question about her street if she hasn't mapped it. Additionally, only the student can let go of an appealing notion when the data contradicts it. One boy from Mumbai told us that giving up on a plan when his own data indicated it wouldn't work was his greatest lesson.

The options could be expanded by the AI. The answer still needed to be earned by the student. And it wasn't the machine that made that possible. The mentor encouraged students to watch, evaluate, and justify their decisions.

This is significant because the change has already started. Students routinely turn to chatbots for explanations and homework help. AI tutors promise every child a patient, personalised guide. And from the 2026-27 session, India has introduced AI into the school curriculum from Class 3 onwards.

There is a great deal to welcome here.

Personalisation has always been the missing piece in Indian classrooms of forty or fifty students. AI can give a struggling student a fifth explanation without impatience, and a fast learner harder problems without waiting for the class. It can take marking and routine practice off a teacher's desk. For students whose families cannot afford private tuition, it may be the first individual attention they receive outside school.

But the risks are just as real.

If AI can complete a worksheet in seconds, the worksheet no longer tells a teacher what a student knows. Homework, as we have designed it, quietly stops measuring learning. Students who outsource the struggle also outsource the thinking that struggle produces. Teachers are being asked to manage AI in classrooms without being trained to use it, question it, or recognise when it is wrong. And if well-off families buy premium AI tutors while under-resourced schools receive tools without trained teachers, we will have automated inequality rather than reduced it.

Here is where the choice sits. Introduced without redesign, AI simply makes an old model more efficient. That is not transformation. That is maintenance.

Redesign means three things. Assessment must move from whether a student can produce an answer to whether she can explain her reasoning, defend a choice and apply an idea somewhere new. Project-based learning must stop being an optional extra, because real, local problems are where AI's limits and a student's judgement both become visible. And AI literacy cannot be a one-day workshop for teachers. It has to sit inside teacher education and continuous professional development, with the time and incentives to make it real.

The National Education Policy already asks for experiential, inquiry-led learning. AI makes that ambition more achievable, and more urgent.

The future of education will not be decided by how intelligent the machines become. It will be decided by whether we let them carry the work teachers were overloaded with, so teachers can return to the work only they can do.

AI can tell a student what the answer is. Only a teacher can help her decide which questions are worth asking.

(The views expressed are personal)

This article is authored by Rahul Ramachandran, director, Partnerships and Initiatives, KRUU.