Future of Education

Personalised Learning at Scale: How AGI Tutors Are Changing Mastery

AGI-class tutors can pair every learner with a system that adapts pace, explanation style, and assessment. The pedagogical and policy implications.

Risograph illustration of a future classroom with AI co-teachers and personalised learning panels — Plate / The classroom is becoming a hybrid human–machine learning environment.

Executive summary

Personalised learning was a promise that schooling could not deliver at scale because it required impossible teacher-to-student ratios. AI tutors remove that constraint. Early evidence is encouraging, but design quality, equity of access, and student motivation determine actual outcomes.

Key concepts

Adaptive pacing
Mastery learning
Spaced repetition
Formative assessment
Motivation and oversight

What 'personalised' actually means

A personalised tutor adjusts at least four dimensions for each learner: pace, explanation style, prerequisite gap-filling, and difficulty. Frontier models can do all four reasonably well today and are improving rapidly.

The mastery effect

Mastery learning — staying on a concept until the learner can demonstrate it — produces large gains when teachers can implement it. They mostly cannot, because a class of thirty makes it impractical. AI tutors restore mastery learning as the default option.

Where AI tutors are weak

Open-ended creative work, group dynamics, and the motivation problem (getting a tired teenager to engage at all) remain hard. AI tutors are not a substitute for the social and emotional scaffolding a good teacher provides.

Implementation lessons

Pilots that work share a pattern: clear learning objectives, teacher-in-the-loop oversight, regular human assessment of original work, and integration into the existing curriculum rather than a parallel track.

Key takeaways

01Personalised learning at scale is now technically possible.
02Gains are strongest in well-bounded subjects with clear objectives.
03Motivation, social learning, and creative work still need humans.
04Implementation quality matters more than the model used.

Frequently asked questions

Do AI tutors widen or close achievement gaps?

Both have happened in pilots. The deciding factors are device access, teacher training, and whether content is available in the learner's first language.

Can students cheat with these tools?

Yes, on take-home assessments. Schools are responding by shifting toward in-person, AI-assisted assessment that probes capability directly.