Research.

The children got there first.

At the beginning of autumn 2026, the grown-ups agreed that schools must be more human. But are the kids way ahead of the game?

The newest insult in primary school playgrounds, according to the Guardian, is “That’s so AI!” Children aim it at anything that looks convincing but turns out cheap: knock-off goods, overblown claims, a parent’s flimsy excuse. It is never a compliment.

The children are onto something the adults spent this month trying to say. In the space of a few days, Sir Anthony Seldon called for “human-rich” schools, the reformer Peter Hyman warned against outsourcing “the very experiences through which we grow”, and even the government declared human understanding “more important, not less”. Everyone agrees where schools must go. Nobody has said what that looks like in a classroom. We can.

The OECD’s new PISA results show why they are worried. UK 15-year-olds rank eighth of 91 education systems for science, and are climbing. Yet they sit in the bottom third of countries for feeling they belong at school, and for believing school has taught them anything useful for a job. Almost everywhere, curiosity is falling and boredom rising. These are not verdicts on children. They are signals from them.

Sir Anthony names one cause: an education built only on “knowledge-rich” teaching is “squeezing out the arts”. AI adds another. PISA finds that students who use chatbots to summarise, research or draft their work score around 20 points lower in science. Nobody can yet say which causes which, but the OECD’s warning is plain: used carelessly, AI becomes a shortcut around the very thinking school exists to build.

Anyone who has worked in a rehearsal room knows what cannot be outsourced. A machine can write a song about a river in seconds. It cannot stand in front of the class and sing it, argue over the second verse, or explain why the chorus changes key. That is the part of learning nobody can do for you.

So, during a normal school week, we put a working artist, a composer, a director or a designer, beside the class teacher. The lesson is still fractions, or rivers. The children do the thinking: together they make something, test it, rework it and explain it. The OECD calls teaching like this cognitive activation, and links it to higher scores and greater enjoyment. When the artist leaves, the teacher keeps the method.

We teach AI the same way. As the OECD’s Andreas Schleicher says, the students who thrive will be those who know when to trust a machine, not those who can merely prompt one. So children set their own work beside a machine’s and argue over which deserves their trust. It is the playground verdict, with reasons.

At the end of a week-long pilot in a primary school, designed and delivered by our team with AI in Education, the charity Sir Anthony co-founded, 83 per cent of pupils said they understood that AI is a tool, not a friend. Their teachers’ confidence in talking about AI rose from 2.7 to 3.6 out of five.

Fifty years ago next month, James Callaghan went to Ruskin College and asked what schools are for. This month, the clearest answer came from the playground. Children already know what fake feels like. Our job is to turn that instinct into judgement, and give every child the chance to make something real.