This talk is an argument for direction over prescription.
I want to start with a very simple question. Think about a child who walked into your school this year. Five years old, maybe six. Can you picture someone specific?
Good. Hold that.
That child will finish school in 2037 or 2038. In a world we cannot fully picture yet – partly because of something that didn’t exist in its current form five years ago.
And we are making decisions about their education right now.
Every curriculum in this building was designed for a world that is already changing faster than curricula can keep up with.
That’s not a criticism of anyone in this room.
It’s just the situation we’re in.
Maps take time to make.
And by the time they’re printed, the territory has shifted.
Something is shifting right now. Faster than anything we’ve seen before.
It’s called AI. And it is not happening to one industry. It is happening to all of them. At once.
What AI can already do — and it’s probably not what you think
When most people hear ‘AI’, they picture robots, or science fiction, or something very technical and far away.
What I want to show you is much more ordinary — and much more close.
I’m going to walk you through five things AI can do today. Right now.
Not in five years. Today.
And as I do, I want you to think about how much of your curriculum — how many hours in the school year — are spent teaching children to do exactly these things.
E X A M P L E 1 — W R I T I N G
AI can write a perfectly competent essay on any topic at secondary school level in about four seconds. Not a brilliant essay. Not one with genuine insight or personal voice. But a passing essay. One that follows the structure, uses appropriate vocabulary, makes a coherent argument.
If I asked it right now to write a 500-word essay on the causes of the First World War, it would hand me something back in moments that would receive a passing grade in most schools.
E X A M P L E 2 — R E A D I N G A N D S U M M A R I S I N G
Feed AI a 200-page textbook chapter. It will give you the key points in seconds, with quiz questions, slides, mind maps etc. attached. It can read a scientific paper and explain it in plain language. It can take a dense legal document and tell you what it actually says.
One of the things we spend enormous time teaching children is how to read closely, identify the main argument, and summarise it accurately. AI does this fluently, in seconds, in any subject.
E X A M P L E 3 — M A T H S
AI solves standard problems. It shows its working. It explains each step. It will tutor a child at two in the morning when the exam is in six hours, with infinite patience, adjusting its explanations until the child understands.
E X A M P L E 4 — L A N G U A G E S
Near-instant, near-fluent translation across more than a hundred languages. Not perfect. But genuinely useful. The kind of translation that would have cost a professional thousands of rupees and several days — done in seconds.
E X A M P L E 5 — C O D I N G
AI writes working computer programmes from plain English descriptions. You tell it: ‘Write me a programme that takes a list of student names and organises them alphabetically’. It does it.
So. I want to stop here for a moment. Because I know what some of you are thinking.
You’re thinking: yes, but AI makes mistakes.
And that’s exactly why this matters.
Because the skill you need — the one that becomes more valuable, not less — is knowing when to trust it, when to question it, when to override it.
But here is the question I want to sit with. Not: should we ban it? Not: how do we stop children using it? Those are important questions, but they are the wrong starting point.
What do we teach now — and what do we stop?
What this means for the jobs these children are aiming for
Let me take you out of the school for a moment and into the workplace. Because ultimately that is what we are preparing children for. And the workplace is changing in ways that most curriculum discussions haven’t caught up with yet.
I want to give you three examples. Not extreme examples. Ordinary, respectable, middle-class careers that parents in your school are probably hoping their children will end up in.
E X A M P L E 1 — T H E P A R A L E G A L
A paralegal spends a significant part of their working day reviewing documents: contracts, court filings, case histories. Flagging relevant clauses. Summarising what matters. Preparing briefings for the lawyers above them.
AI does this now. Not perfectly. But accurately enough that law firms are already using it. The paralegal job is not gone. But it has shrunk. Fewer people are needed to do the same amount of document review. The ones who remain need different skills than the ones who trained five years ago.
E X A M P L E 2 — T H E J U N I O R D O C T O R
A junior doctor reads scans. Interprets results. Writes referral letters. Follows diagnostic protocols.
AI diagnostic tools are already more accurate than junior doctors on certain imaging tasks — detecting early signs of certain cancers in scans, for example. The doctor is not being replaced. But the role is changing. The entry-level work — the pattern recognition, the protocol following, the document writing — that is increasingly automated.
Which raises a question nobody in medical education has yet fully answered: how do you train a doctor if the first five years of practice — the years where you learn by doing the routine work — no longer exist in the same form?
E X A M P L E 3 — T H E A C C O U N T A N T
Standard tax returns. Financial summaries. Compliance checks. Spotting anomalies in spreadsheets.
These are the things graduate accountants spend their first years doing. Learning the craft by doing the work. AI now does much of it faster and with fewer errors.
The accountant who succeeds in 2038 is not the one who can process a tax return correctly. That’s table stakes. It’s the one who can ask the right question of the data. Who can sit with a client who is frightened about their finances and help them think. Who can spot the pattern that matters, not just the anomaly the algorithm flagged.
T H E P A T T E R N U N D E R N E A T H A L L T H R E E
It is not that the jobs vanish overnight. It is subtler than that, and in some ways more disruptive.
The entry level disappears. The first five years of a knowledge career — the years where you learn by doing the routine work — those are being automated.
Which means: what does that do to the pipeline?
How do people develop into senior lawyers, experienced doctors, trusted accountants, if the junior work that used to train them no longer exists in the same form?”
What are we training these children to be? Because ‘junior anything’ — junior analyst, junior writer, junior researcher — is exactly what AI is eating first.
And this is exactly what most schools are optimised to produce.
The entry-level jobs being automated are, almost without exception, compliance jobs: follow the protocol, apply the rule, produce the standard output.
What AI cannot automate is the person who questions whether the protocol is right.
Schools that reward compliance above all else — the right answer, the correct format, the expected response — are training children for the jobs that will disappear fastest.
The mistake most people make looking at AI’s impact is to look at individual jobs: will a paralegal be replaced? Will a junior doctor?
But the real disruption is systemic.
It is the whole pipeline of how people develop in a profession. When you look at the object (the job) you miss the system (the career pathway, the training structure, the apprenticeship model).
That is what is actually changing.
And nobody has yet worked out what replaces it.
So what does school actually need to become?
I am not going to stand here and tell you I have the answer.
Nobody does.
Anyone who says they know exactly what school should look like in 2038 is either very confident or not paying close enough attention.
What I can offer you is a different question to start from.
Not: ‘do we teach AI?’
Not: ‘how do we incorporate AI tools into the curriculum?’
BUT:
What becomes more valuable when AI handles the routine?
Because something does become more valuable. Several things, in fact.
1 — J U D G M E N T U N D E R U N C E R T A I N T Y
AI gives you the best statistical answer. It is extraordinarily good at that. What it cannot do is sit with ambiguity. Weigh competing values. Make a call that a community will trust — not because it’s mathematically optimal, but because a human being with judgment and accountability made it.
That is a skill. It can be taught. It requires practice, complexity, and real consequences — even small ones. It does not develop through multiple choice questions with one correct answer.
Schools barely teach it. Most curricula actively work against it, because judgment is hard to standardise and even harder to grade.
Judgment cannot be installed from above. You cannot teach judgment by telling children what the right answer is — that produces the opposite.
Judgment develops when children are put in situations where the answer is genuinely unclear, where their own thinking matters, where they can see the consequences of their choices.
Your job as a teacher is not to deliver the correct judgment — it is to create the conditions in which judgment can develop.
Notice who steps forward. Notice what emerges when you stop providing the answer.
2 — G E N U I N E C U R I O S I T Y A N D T H E A B I L I TY T O A S K A B E T T E R Q U E S T I O N
AI answers questions magnificently. It is genuinely bad at knowing which question matters.
The child who asks ‘why does this rule exist, and what happens at its edges?’ is asking something more valuable than the child who can correctly apply the rule. The researcher who asks an unexpected question is more valuable than one who efficiently answers an expected one.
Schools tend to reward the right answer. They grade outputs. They measure whether you knew the thing you were supposed to know. That’s understandable — it’s easier to assess. But it trains children for a world where the premium is on answers, when the world they’re entering puts the premium on questions.
The child who generates their own question owns it in a way that the child who answers your question never does. What people create themselves, they maintain.
Curiosity cannot be pushed into a child — it has to be pulled out by creating the right conditions: open problems, real stakes, genuine room for the child’s own thinking to matter.
Ask yourself: in your classroom right now, who is generating the questions? You, or them?
3 — R E L A T I O N S H I P S A N D T R U S T
A doctor who can communicate with a frightened family. A teacher who notices a child is struggling before any test shows it. A manager who holds a team together through uncertainty, not because they have the answer but because people trust them.
These are not soft skills. They are the skills that will define whether a person is valuable in a world where AI handles the technical work.
They are profoundly undertaught in most schools, because they do not appear on any exam paper. And schools teach what they measure.
O N E C O N C R E T E T H I N G T O T A K E B A C K
I don’t want to leave you with three big ideas and no action. So here is the most practical thing I can offer.
Stop grading outputs. Start grading process.
Ask: What did this child notice?
What did they try when the first approach didn’t work?
What question did they ask this month that they couldn’t have asked last month?
What changed in how they think?
This is not theoretical. Riverside School in Ahmedabad has been doing this for twenty years. Their students are assessed not on what they remembered but on what they changed.
What problem did you identify?
What did you try?
What happened?
What did you learn from what didn’t work?
Parents see that portfolio. They understand it. Because it tells them something a grade never could: who their child is becoming.
That’s the curriculum that survives AI.
Not because it’s clever or modern.
Because it develops the thing AI cannot replicate: a human being who is actually growing.So what does school actually need to become?
I am not going to stand here and tell you I have the answer.
Nobody does.
Anyone who says they know exactly what school should look like in 2038 is either very confident or not paying close enough attention.
What I can offer you is a different question to start from.
Not: ‘do we teach AI?’
Not: ‘how do we incorporate AI tools into the curriculum?’
BUT:
What becomes more valuable when AI handles the routine?
Because something does become more valuable. Several things, in fact.
1 — J U D G M E N T U N D E R U N C E R T A I N T Y
AI gives you the best statistical answer. It is extraordinarily good at that. What it cannot do is sit with ambiguity. Weigh competing values. Make a call that a community will trust — not because it’s mathematically optimal, but because a human being with judgment and accountability made it.
That is a skill. It can be taught. It requires practice, complexity, and real consequences — even small ones. It does not develop through multiple choice questions with one correct answer.
Schools barely teach it. Most curricula actively work against it, because judgment is hard to standardise and even harder to grade.
Judgment cannot be installed from above. You cannot teach judgment by telling children what the right answer is — that produces the opposite.
Judgment develops when children are put in situations where the answer is genuinely unclear, where their own thinking matters, where they can see the consequences of their choices.
Your job as a teacher is not to deliver the correct judgment — it is to create the conditions in which judgment can develop.
Notice who steps forward. Notice what emerges when you stop providing the answer.
2 — G E N U I N E C U R I O S I T Y A N D T H E A B I L I TY T O A S K A B E T T E R Q U E S T I O N
AI answers questions magnificently. It is genuinely bad at knowing which question matters.
The child who asks ‘why does this rule exist, and what happens at its edges?’ is asking something more valuable than the child who can correctly apply the rule. The researcher who asks an unexpected question is more valuable than one who efficiently answers an expected one.
Schools tend to reward the right answer. They grade outputs. They measure whether you knew the thing you were supposed to know. That’s understandable — it’s easier to assess. But it trains children for a world where the premium is on answers, when the world they’re entering puts the premium on questions.
The child who generates their own question owns it in a way that the child who answers your question never does. What people create themselves, they maintain.
Curiosity cannot be pushed into a child — it has to be pulled out by creating the right conditions: open problems, real stakes, genuine room for the child’s own thinking to matter.
Ask yourself: in your classroom right now, who is generating the questions? You, or them?
3 — R E L A T I O N S H I P S A N D T R U S T
A doctor who can communicate with a frightened family. A teacher who notices a child is struggling before any test shows it. A manager who holds a team together through uncertainty, not because they have the answer but because people trust them.
These are not soft skills. They are the skills that will define whether a person is valuable in a world where AI handles the technical work.
They are profoundly undertaught in most schools, because they do not appear on any exam paper. And schools teach what they measure.
O N E C O N C R E T E T H I N G T O T A K E B A C K
I don’t want to leave you with three big ideas and no action. So here is the most practical thing I can offer.
Stop grading outputs. Start grading process.
Ask: What did this child notice?
What did they try when the first approach didn’t work?
What question did they ask this month that they couldn’t have asked last month?
What changed in how they think?
This is not theoretical. Riverside School in Ahmedabad has been doing this for twenty years. Their students are assessed not on what they remembered but on what they changed.
What problem did you identify?
What did you try?
What happened?
What did you learn from what didn’t work?
Parents see that portfolio. They understand it. Because it tells them something a grade never could: who their child is becoming.
That’s the curriculum that survives AI.
Not because it’s clever or modern.
Because it develops the thing AI cannot replicate: a human being who is actually growing.
What school can do — right now
You are going to walk back into a school tomorrow where children have to pass exams. Real exams, with real grades, with real consequences for their futures.
I am not going to pretend that constraint does not exist. It does. And any school that ignores it is failing its students in a different way.
So the question is not: how do we abolish exams and start again?
The question is:
what can you change — now, this term, without a new budget or a ministry of education circular — that moves in the right direction while still getting the children through?
Preparing children for their exams and preparing them for their lives are not opposites. They never were.
Let me show you what I mean.
P R I M A R Y S C H O O L ( A G E S 5 – 1 1 )
In primary, the curriculum pressure is real but the shape of the day is still largely in the teacher’s hands.
The most powerful changes here cost nothing.
- Start every topic with a question, not an answer
Not: “Today we are going to learn about rivers.” But: “Why does water always end up in the sea?” Same content. Same curriculum standard. Completely different relationship with the knowledge.
The child who first heard a question is looking for an answer.
The child who first heard an answer is waiting to be told what to remember.
Those two children will handle AI very differently in ten years. - Make failure visible and normal
In Janwaar we had a rule: you fall off the skateboard, you get back on. Nobody laughs, nobody rescues you. You figure it out. That sounds like a lesson about skateboarding. It is actually a lesson about learning.
Primary teachers: when a child gets something wrong in class, your instinct is to move quickly to the right answer. Try instead: “Interesting. Why did you think that?” Four words. They do not cost curriculum time. But they teach something AI cannot: that your thinking process matters, not just your output. - Give children real problems with no pre-agreed answer
Once a week. Fifteen minutes. A real problem from the school’s actual life:
The lunch queue takes too long — how do we fix it?
The corridor outside is too noisy at lunch — what should we change?
The library has money to spend — what should it buy?
Not hypothetical. Real. With real consequences. That is the difference. Children know immediately when a problem has been invented for them to practise on. They engage completely differently when they know something will actually happen as a result of their thinking.
Real problems with real stakes pull children into genuine thinking. Textbook exercises push content at them.
Same subject, completely different quality of engagement — and completely different preparation for a world where the questions matter more than the answers.
S E C O N D A R Y S C H O O L ( A G E S 1 1 – 1 8 )
In secondary the exam pressure is harder and more explicit. But the students are also older — which means they can handle more open conversations. And open conversations turn out to be one of the most powerful things you can offer them.
- Tell them the truth about the exam — and what is beyond it
Try saying this to a class of fifteen-year-olds: “The exam you are preparing for tests whether you can remember and reproduce certain things under pressure. That is a real skill and it matters. I am going to help you pass it. And I am also going to help you develop skills the exam does not test — because those are the ones that will matter more when the exam is over.”
Most teenagers have never been told that by a teacher. The effect is remarkable. They understand that you are taking them seriously. They also start to understand that school is not the only thing happening to them — that there is a larger project underway, and they are part of it. - Use AI in the classroom — critically, not passively
Ask AI to write an essay on your topic. Then give it to the class and ask: what is wrong with this? What is missing? What would a human writer know that this does not? What question did it not think to ask?
This does three things at once. It teaches the subject — students read and analyse the content. It teaches critical thinking — they have to evaluate and find the gaps. And it teaches AI literacy — they stop seeing it as an oracle and start seeing it as a tool with limitations. That is a skill they will use for the rest of their lives. - Build at least one thing across the year that no exam will assess
One project, per class, per year. Something real. A product, a campaign, a service, a piece of research that goes somewhere beyond the classroom wall. It could be tiny — a school newspaper, a garden, a presentation to the local council, a survey of the neighbourhood that actually informs something.
The point is not the output.
The point is that the children experience what it feels like to make something — to take responsibility for it, to navigate the gap between intention and reality, to finish it.
No exam teaches that.
No AI can replicate the experience of having genuinely built something yourself.
F O R S C H O O L L E A D E R S S P E C I F I C A L L Y
Everything above is in the hands of individual teachers. But leaders create the conditions. Three things that cost almost nothing but change everything:
- Protect the time for things that CANNOT be measured
In every school I have visited, the things that develop judgment, curiosity and character — the conversations that go somewhere unexpected, the projects that take on a life of their own, the moments when a teacher follows a student’s line of thinking instead of the lesson plan — these are the first things cut when time pressure mounts.
Protecting that time is a leadership act. It requires saying, explicitly and repeatedly: this matters, even though it will not appear on any spreadsheet. That is harder than it sounds.
But it is one of the most important things a school leader can do right now. - Stop rewarding the teachers who produce the highest scores
Or rather — stop rewarding only those teachers.
In most schools, the visible markers of a good teacher are exam results, compliance with the lesson plan, and absence of problems. None of those things are wrong to track. But they are insufficient.
Start asking a different question in appraisals:
What did your students try this year that they were not sure would work?
If no one can answer that, something is missing. The teacher who produces curious, questioning, slightly difficult students is doing something at least as valuable as the one who produces high-scoring compliant ones. - Ask the students what they think school is for
Not in a survey. In a real conversation, with real consequences. Tell them: we are thinking about how our school should change. We want to know what you think. And then — actually do something with what they say.
The students who are asked what school is for, and who see their answers taken seriously, are already practising something AI cannot replicate: participation in the design of their own environment. That is not a nice extra. It is the thing itself.
None of this requires a new curriculum.
None of it requires a budget.
It requires a decision about what kind of school you want to be.
And the courage to start being it.
I want to end not with a prediction, but with what I think is the honest shape of what we are facing.
The biggest disruption AI brings to education is not to the curriculum. It is not about which subjects to teach or which tools to allow in exams.
Those are real questions, but they are downstream of something bigger.
For a hundred years, school has been the place where knowledge lives. Where you go to get it. The teacher knows. The textbook has it. You come to school to receive it.
AI ends that.
Knowledge is now everywhere, instantly, for free. Any child with a phone can access more information in ten seconds than a library held in 1990. And they can have it explained to them, patiently, at whatever level they need, at any time of day.
Which means school has to become the place where something else happens. Something harder to automate. Something more human.
What that is — exactly — I think is the most important question in education right now. And I don’t think it gets answered by policy, or by curriculum committees, or by AI companies.
I think it gets answered by people like the ones in this room. By teachers and school leaders who are paying attention. Who are willing to ask uncomfortable questions about what they are actually doing and why.
The children starting school today will graduate into a world we cannot fully picture. They are counting on you to think carefully about what they actually need.
Not what the system requires.
What they need.
Thank you.