■ MODERATE RISK ■ Education
No — but the job changes shape, and the change is arriving faster than most school systems are ready for. Instruction can be personalized by machine; classroom management, safeguarding, and childhood development cannot.
“Kids need role models, not chatbots. AI tutors the facts; teachers build the humans.”
Our AI replacement risk score — how we score jobs
Teaching a class of thirty is only partly about content. A teacher plans lessons against a curriculum, delivers them at a pace that works for a room of wildly different abilities, marks work, tracks progress, manages behaviour, spots the child who came in hungry or bruised, communicates with parents who range from absent to litigious, sits through professional development, and completes an administrative load that most people outside the profession badly underestimate.
AI addresses the content and admin layers well. Lesson planning, differentiated worksheets, quiz generation, marking of structured work, report card comments, and translation for families whose first language differs all compress dramatically — genuinely useful, since teacher workload is a leading cause of people leaving the profession. Adaptive tutoring systems can drill maths or vocabulary with per-student pacing that no human can sustain across thirty children simultaneously. For motivated older students, an AI tutor is a real substitute for some direct instruction, and homework in its traditional form is already broken.
What resists is nearly everything else. Children don't self-direct; a room of eight-year-olds with tablets and no adult produces chaos, not learning. Motivation, attention, and behaviour are managed through relationship — a child works hard for a teacher who notices them. Safeguarding is a legal and moral duty requiring an adult who sees the child daily. Socialization, conflict resolution, and the modelling of adult behaviour are the hidden curriculum that schools actually deliver. Special educational needs support demands human adaptation and, often, physical care. And schools are custodial institutions: parents need somewhere their children safely are. Our 30 says AI becomes a powerful teaching instrument, not a teacher.
Automatability: our editorial assessment of current and near-term AI capability
Change is arriving as workload relief first: planning and marking tools are spreading through schools now and will be near-universal by 2030. By 2040 the classroom likely looks different — more AI-driven personalized practice, more teacher time on coaching, discussion, and pastoral work. Teacher numbers are driven far more by demographics and funding than by automation, and most systems currently face shortages rather than surpluses.
No, and the reason is less about instruction than about children. Learning at school depends on motivation, attention, behaviour, and safety — all of which are managed through an adult relationship. AI is a strong tutoring and workload tool, but a class of children with software and no teacher doesn't produce education. Our risk score of 30 reflects a job transformed, not removed.
For focused, well-defined practice with a motivated learner, it's genuinely excellent — infinitely patient, always available, and paced to the individual. For a mixed-ability class of thirty children of varying enthusiasm, it isn't a substitute at all. The realistic best outcome is a teacher who uses adaptive practice tools to free up time for the pupils who need direct attention.
Fastest at the desk, not the front of the room. Lesson plans, differentiated worksheets, quizzes, marking of structured tasks, and parent communications all get much faster. It's also forcing assessment redesign, because homework essays can be generated. The direction of travel is less preparation and paperwork, more direct interaction, discussion, and pastoral work.
Yes, if the pastoral and relational side appeals — that's where the durable value sits. Most education systems face shortages driven by workload and pay rather than any oversupply, and AI plausibly improves the workload problem. What's changing is the skill emphasis: less time as a content deliverer, more as a coach, behaviour manager, and designer of learning experiences.