Coaching produces the largest average gains of any model of teacher professional development.
Two major reviews of the causal evidence say the same thing. Coaching improves what teachers do in the classroom, and those improvements reach pupils. The question is not whether coaching works. It is whether schools can afford or manage to offer it at scale.
Why coaching works where training days don't.
Professional development conferences, training days, and whole-school INSET have been the default for teacher development for fifty years. The evidence is consistent: they produce little measurable change in classroom practice. Darling-Hammond et al. (2017)
Coaching works because it is tied to what is actually happening in a teacher's classroom. Observation, feedback, deliberate practice, and refinement — repeated over time, focused on one specific aspect of practice. It is the same mechanism that produces expertise in surgery, sport, music, and every other demanding skilled profession.
A 0.49 standard deviation effect on instructional practice is large. For context, most professional development interventions produce effects close to zero.
If the evidence is this clear, why doesn't every teacher have access to a great coach?
Because traditional coaching — expert, sustained, one-to-one, in-person — runs into six structural barriers. These are not problems of will or effort. They are systemic constraints that no amount of good intention has ever resolved.
Not a better way to manage coaching — the coach itself.
Most responses to the scaling problem try to make human coaching more efficient — better frameworks, structured cycles, digital tools, dashboards for coaches to use. The coaching relationship still depends on a human coach being available, trained, consistent, protected from role drift, and clear of every one of the six barriers above.
Aristotal takes a different position. The AI is the coach. That is not a marginal improvement on the existing model. It is a structural response to each of the six barriers — simultaneously.
What AI-as-coach makes available that no other model can.
Removing the six barriers is not only a cost and logistics story. It also makes possible three things that traditional coaching — however well designed — cannot deliver at the scale of a whole school or trust.
Homewood School & Sixth Form Centre, Kent.
Homewood had already built one of the most systematic CPD programmes in the country — team coaching, instructional coaching frameworks, individual coaching software. Paul Hanson, Assistant Vice Principal for Staff Development and Research, had done more than most school leaders to close the coaching gap. Two things remained out of reach: short feedback loops, and equal provision for every teacher.
“Knowing that the feedback is going to be consistent, high quality and evidence informed — I could not guarantee that with every human coach.”Paul Hanson · AVP Staff Development & Research · Homewood School
“Just having a transcript — which you don't even need to click into — was 100% the game changer. For the first time, you've suddenly got 97 staff recording themselves. That's just a sea change.”Paul Hanson · AVP Staff Development & Research · Homewood School
“My disruption has about halved and my stress management has got much better because I am not fighting against my own personality.”Secondary Teacher · Homewood School and Sixth Form Centre
The same pattern, in different schools.
Best AI Tool for Teaching and Assessment — the largest education technology event in the world.
Aristotal's work is guided by Dylan Wiliam, Emeritus Professor, UCL Institute of Education, one of the most cited researchers in formative assessment and teacher professional development; Mary Myatt, writer and curriculum consultant whose work on high-challenge, low-threat learning is used in schools across the UK; and John Tomsett, headteacher and author whose practice-based writing on teacher development has shaped a generation of school leaders.