By: Saunak Ghosh, Co-Founder & CEO, TalentGum
A child today can ask an AI tool almost anything – why the sky is blue, how a rocket works, what caused a war centuries ago – and get a clear, well-structured answer in seconds. It is, in many ways, remarkable. Knowledge that once took a trip to the library, or a patient adult willing to explain, is now available instantly, endlessly, on demand. But somewhere in that convenience sits a harder question that parents and educators are only beginning to grapple with: if a machine can answer almost anything, what exactly are children learning to do themselves?
Real thinking was never really about having access to information. It was about what happens between the question and the answer – the noticing, the guessing, the getting it wrong, the trying again. Picture two children solving the same maths problem. One struggles with it for ten minutes before finally cracking it. The other types it into an AI tool and has the answer in three seconds. The first child walks away with a method they can reuse the next time a problem looks unfamiliar, and perhaps a little more resilience than they started with. The second walks away with the correct answer, but not necessarily the ability to solve the next unfamiliar problem. AI is extraordinarily good at shortening, or even erasing, that middle stretch of struggle. And that middle stretch is precisely where thinking happens.
None of this is an argument against AI. Telling children to avoid tools that are becoming basic infrastructure of the world would be about as realistic as telling a child a generation ago to avoid calculators, or the internet. The question was never whether children would grow up surrounded by AI – they will. The real question is what children will have learned to do independently, and how confidently they will be able to think for themselves, before AI becomes part of how they learn and work.
It helps to draw a line between AI used as a shortcut and AI used as a sparring partner. As a shortcut, the pattern is simple: a child asks, the tool answers, the thinking stops. As a sparring partner, the pattern has more steps: a child asks, gets a response, and then does something with it – questions it, tests it, pushes back, asks a follow-up that goes a layer deeper. Notice that the tool itself doesn’t change between these two scenarios. What changes is the habit the child brings to it. And habits like that aren’t automatic; they’re built over time, through repeated experiences of being encouraged to question, experiment, and get things wrong without it being a disaster.
This is where parents and educators have real leverage – not by restricting usage of technology but by shaping the environment around it. A child who has regularly been asked to explain their reasoning, defend a position, or work something out without being handed the answer develops a fundamentally different relationship with information than a child who hasn’t had that practice. When that child eventually sits down with an AI tool, they are far more likely to treat its answer as a starting point rather than a finish line – simply because they have already spent years practising the habit of pushing past a first answer, long before any AI tool ever gave them one instantly.
At TalentGum, this shows up directly in how we design our classes: the goal is never simply to hand a child the right answer, but to build the confidence and ability to work toward one. A chess class only works if a child sits with a difficult position, weighs several options, and figures out why one move beats another. A public speaking session doesn’t build confidence by handing a child a polished script; children build confidence when they structure their own argument, stumble through a first draft, and sharpen their delivery over multiple attempts with feedback. Even our AI & Machine Learning course, ironically, is built the same way: rather than letting children treat AI as a black box that gives them answers, we have them build small models and projects from scratch – training a classifier, testing where it fails, figuring out why – so they come away understanding how the “magic” actually works instead of just consuming it. That’s also what live, curriculum-led teaching adds that a search box can’t: a teacher who can watch how a child is thinking, ask “why do you think that?”, and nudge her to refine an idea rather than just deliver the correct one.
That kind of guided learning only becomes more valuable as answers get easier to obtain. The goal was never to out-race AI on speed of information delivery – that’s a race no human will win. The goal is to help children use information well once they have it: to interrogate an answer, look at a problem from a different angle, spot when something doesn’t add up, and decide what to do next. None of that comes from collecting more answers. It comes from someone challenging a child’s thinking and then giving them room to work through it themself..
For parents, none of this requires a rulebook. Try a simple experiment over the next week: notice how often your child reaches for an instant answer versus how often they sit with a question for a while first. Neither instinct is wrong on its own – plenty of questions genuinely just need a quick, correct answer and nothing more. But if the instant answer has quietly become the default for almost everything, that’s worth paying attention to. The muscle for sitting with uncertainty, for working something out slowly, atrophies exactly like any other muscle left unused.
AI can hand a child an answer in seconds. What it cannot hand them is curiosity, patience with difficulty, or the judgment to know when an answer deserves more scrutiny. Those are still built the old way – through practice, through struggle, through adults and mentors who choose to ask one more question instead of supplying one more answer. The tools our children grow up with will keep changing faster than most of us can track. What won’t change is that thinking is still something children have to do themselves, one hard problem at a time – and the best thing we can do is make sure they get enough chances to do exactly that.
