AI & the future of education

An AI tutor is a language model set up to teach rather than to answer. It asks what the pupil already knows, offers a hint, shows an intermediate step and then waits. The thinking stays with the pupil, and that is where the learning happens. In a controlled trial in a large physics course, students working with such a tutor learned more than twice as much, in less time, as in a class built on active learning. Large language models now have a learning mode of their own, and they can already speak and use a camera to see what a pupil is working on. The next step is a tutor that uses this as a teacher would: it sees the arithmetic slip in a pupil's exercise book and responds with a question before the mistake becomes a habit.

While pupils practise, adaptive platforms record which mistakes keep coming back and how long each exercise takes, and choose the next exercise accordingly. Over weeks and months, the teacher builds up a picture that also includes pupils whose passing marks hide their difficulties. The next step is for that information to reach beyond today's topic. A gap in fractions sometimes only shows up in algebra. The platform estimates earlier which topics further on may prove difficult and adjusts the revision to match. Today's exercise then doubles as preparation for next year's material.

Lesson preparation takes up much of teachers' time outside lessons, and AI now does far more than write a first draft. A teacher describes the aim and the group, and receives a lesson with exercises, a case from the pupils' own surroundings and a short simulation, at three levels. Linked to the class's results, the material concentrates on what pupils have not yet mastered. In a large survey of teachers, weekly AI users estimated their combined time savings across various teaching tasks at almost six hours a week on average. Future material could adapt while the lesson is running, using the mistakes of the first half hour to shape the exercises in the second.

Much of the time in education goes on administration and questions that come back every term. When is the resit? Which documents are still missing for enrolment? An AI agent with access to the student system and knowledge of the regulations answers those questions day and night and handles simple requests itself: looking up records, filling in a form, booking an appointment. Universities and colleges are trialling such assistants for student services, and schools let AI help with the timetable. Staff deal with the exceptions. The next step is an AI agent that guides a student through the whole programme, from application to graduation, and sees for itself when a resit, an exemption or a different module would help them forward.

AI literacy is the ability to understand AI, use it critically and grasp what follows from it. Critical thinking sits at its core. A language model predicts words and can therefore write convincing nonsense; a photo, video or voice may be fabricated. So a child or student learns to ask questions: is this right, where does it come from, who gains from it, and what does another source say? UNESCO has drawn up frameworks for what pupils, students and teachers should be able to do, and governments are adding AI literacy to the curriculum. The European AI Act requires organisations that use AI, schools included, to take steps to support the AI literacy of their staff. The OECD will measure AI literacy in its 2029 PISA study.

A newly arrived pupil reads the same source as the rest of the class, with the instructions in their mother tongue alongside. A student who is hard of hearing follows the discussion through live subtitles, a blind student has a graph described, and a pupil with dyslexia has the test read aloud in a natural voice. AI builds that support into the teaching material itself: a text gets several reading levels, an image a description, a spoken explanation a written version. The whole class works on the same assignment, each with their own way into the material. The next step is sign language. Researchers are working on AI that translates sign language live, in both directions, so that a deaf pupil and a hearing classmate can talk to each other directly.

A student still passes, but hands work in later and later, misses lectures and logs in to the learning platform less often. On their own, such signals say little; together they form a pattern that becomes visible before the grades drop. A predictive model learns from earlier years which combinations of signals often come before a student drops out. Students with a raised risk appear on a list for their personal tutor, who gets in touch, asks how things are going and looks at what would help: extra tuition, an adjusted study plan, support with money worries. The model gives a probability, not a certainty. The conversation that follows is what makes the difference.

A biology pupil walks through a cell, a vocational student practises welding without wasting material, a nursing student breaks bad news to a virtual patient. AI makes these simulations smarter. It steers how the virtual people behave, responds to what the student says and adjusts the difficulty: more time pressure for those who are doing well, more help for those who get stuck. Afterwards the student gets feedback on what they did. Meanwhile the headsets are becoming lighter and flatter, and fit more easily into an ordinary classroom.

When AI joins the work, what teachers and support staff do changes. Learning coaches, instructional designers, teacher-researchers and AI coordinators already work in many schools. AI will change what some roles involve and make others far more prominent. The learning architect designs learning routes across schools and employers, so that a vocational student can carry on learning with an employer without a detour. The learning companion supervisor keeps an eye on what AI recommends to a pupil, and steps in when a recommendation does not fit what that pupil needs. The world builder creates simulations and virtual worlds for a subject, from a historic harbour to an operating theatre. Timetablers and administrative staff get different work too, with less data entry and more investigating.

A classroom where teacher and AI work together divides the work according to who is good at what. People are good at the human side: judging, choosing, motivating a pupil. AI takes the repetitive and the deeply analytical work: it sets practice, explains the same topic a fifth time in a different way and spots, across thirty pupils' answers, which misunderstanding keeps returning. The teacher sees what AI does not: the pupil having a hard week at home, the group that is not working well together, the talent nobody had noticed yet. The picture for 2030: at the start of the lesson, the system shows the teacher which three pupils need their attention today, and why. How to respond remains the teacher's call.

School leaders lead the way with AI and new technology, even when not everything is settled. They sketch a picture of the future that a teacher of eight-year-olds and a university lecturer can both understand: where the school uses AI, and where it deliberately chooses not to. They start with small, visible steps, in one department or one programme, and show what works. Teachers get room to experiment and to bring colleagues along. That is how leaders get their people building the school of tomorrow together.