If AI can give the answer, what is school for?

If a child can ask AI to balance a chemical equation, why should a school still teach it?

I do not think the arrival of AI makes school unnecessary. But it does challenge a version of schooling in which producing the correct answer is treated as the end of learning.

The answer may come from a machine. The understanding still has to develop in the child.

If we look back to 1930's to 1960's, a man without a hat out in open would have been a really odd thing to look at - Yet today none wears it. That is because a behaviour can look permanent because it fits the environment that produced it. Then the environment changes in small ways over time. Nothing appears decisive on its own. Yet one day the old behaviour no longer makes sense, even though it seemed completely normal until yesterday.

The hat was my way of making a broader point: systems do not always disappear because somebody formally replaces them. Sometimes the conditions that kept them useful gradually weaken, and behaviour then changes much faster than the environment did.

I think that is the more serious challenge AI presents to school. The classroom, book, notebook and board will not become irrelevant merely because a new tool exists. But if their main job is to deliver and reproduce answers that a child can obtain instantly elsewhere, the environment around them has changed. Their continuing value has to come from work the answer alone cannot complete: building knowledge, testing explanations, applying ideas and forming judgement.

That is not an argument for banning the machine. The OECD's January 2026 review distinguishes better work produced with AI from actual learning, while also finding that thoughtfully designed AI use can support learning.^[1] The question is what we want the child to gain from the work.

Some school models are already making more room for application. Anandi in Bengaluru describes four academic days and a fifth devoted to a semester-long project. SOVO in Mohali describes a similar arrangement.^[2][3] These are the schools' published designs, not independently verified results.

What interests me is not simply the project day. It is whether the knowledge acquired during the other four days becomes something a child can use.

Keep the equation. Change what we do with it.

Consider an illustrative chemistry lesson.

A teacher presents a reaction in an open container. Gas escapes, and the measured mass falls. Yet the class has learned that matter is conserved.

Where did the missing mass go?

The Royal Society of Chemistry uses precisely this distinction between open and closed systems to connect observable changes with conservation of mass and balanced equations.^[4] The equation is not an arbitrary hurdle. It represents the idea that atoms are rearranged, rather than created or destroyed, in the reaction.

A child could memorise the rule and still struggle to explain the observation.

So I would want the lesson to move between the two. Predict what will happen. Examine the result. Explain why the apparent loss does not contradict the principle. Then consider what changes if the gas is retained and included in the measurement.

The point is not to replace chemistry with an entertaining activity. It is to make the chemistry necessary to understanding the activity.

Of course, AI could also explain the escaping gas. That is the strongest challenge to a simple “facts are out, concepts are in” argument: a machine can supply a conceptual explanation too.

But receiving that explanation does not settle whether the child understands it. I would want the child to connect the explanation to the evidence and reason through a changed situation—not merely repeat a more sophisticated answer.

This is also why I would resist the suggestion that children need less knowledge because AI has so much of it.

The National Academies' *How People Learn II* explains how existing knowledge supports reasoning and further learning.^[5] In our example, a learner needs some grasp of atoms, gases and measurement to make sense of the explanation. “Critical thinking” cannot substitute for everything there is to think with.

The educational choice is therefore not knowledge or application. It is whether knowledge remains something to reproduce or becomes something to reason with.

A project is not automatically an education

There is a fair concern here. Will more projects simply mean less time for serious teaching?

Not necessarily. XP School in Doncaster, England describes projects mapped to subject content and academic standards. Arbor School in Dubai combines the English National Curriculum with interdisciplinary projects and direct teaching.^[6][7] Their descriptions show that instruction and application need not be opposites; they do not establish superior outcomes.

The research makes that qualification important.

A study published online in November 2022 tested a project-based science programme with 2,371 third-graders across 46 Michigan schools. Pupils receiving the programme scored higher on a science test by 0.277 standard deviations—just over a quarter of a standard deviation, not 27.7% more learning. The programme included curriculum materials, teacher professional learning and assessments.^[8]

An English Year 7 trial conducted during 2014–16 found no clear benefit for literacy or engagement. Substantial school dropout from the trial made those findings less secure.^[9] These were different programmes and outcomes, not competing versions of the same experiment.

So the evidence does not justify replacing lessons with projects indiscriminately. It supports examining what pupils are actually learning through a particular design.

Imagine our chemistry class producing a beautiful presentation about conservation of mass, assembled largely by AI. We might have replaced a copied worksheet with a polished presentation without changing who did the thinking.

The useful project is the one that makes the child grapple with the idea. Sometimes that will require an experiment. Sometimes a teacher's explanation, a worked example or more practice will be the next useful step.

For an Indian school, I would be more interested in that connection than in a promise of fewer books or more outdoor time. A project day disconnected from the academic week could become another compartment. The opportunity is to let a lesson inform an investigation, and let the investigation reveal what the next lesson needs to address.

That changes the work of the school without making the teacher, classroom or notebook redundant. In the lesson I am proposing, the teacher connects ideas and observations; the notebook records predictions, evidence and revised explanations.

I would not claim that every conventional school will survive unchanged, or that these selected examples prove a worldwide transition. But AI's ability to generate an answer does not make this educational work unnecessary.

My conclusion is that the school worth building for an AI world is not one where children know less. It is one where they learn to do more with what they know, including judging the help a machine provides.

What should become obsolete is not the chemical equation. It is mistaking the reproduction of an answer for an education.