It differs from general-purpose AI use because output is anchored to two things: an explicit instructional model, and a named curriculum reference. The teacher edits and signs off, and the system records that they did — which is what makes it defensible under both DfE guidance and inspection.
The problem this solves
Teachers are already using generative AI. Ofcom research has found substantial and growing AI use among children, and surveys consistently find teacher use running well ahead of school policy. The realistic question is not whether AI enters lesson planning but whether it does so in a form anyone can check.
An unstructured chatbot produces a lesson plan that is fluent, well-formatted, plausible and disconnected from the curriculum objective the teacher is accountable for. It frequently sits six activities at the same cognitive level and includes no check for understanding. It reads better than what it replaced and teaches no better.
DfE guidance and Ofsted’s own research on AI converge on the same principle: staff may use AI, but the school remains responsible for professional judgement, safeguarding, data protection and curriculum quality. Pedagogy with AI is that principle expressed as a product constraint rather than a policy sentence.
Definition
Pedagogy with AI is the use of artificial intelligence to plan, adapt and evidence teaching while the teacher retains professional judgement and accountability. Three conditions distinguish it:
- Model-anchored. Output is generated against a stated instructional model — Bloom’s taxonomy, Rosenshine’s principles, cognitive load theory, the 5E model — not a free-text prompt.
- Curriculum-anchored. Output maps to a specific curriculum objective, so a lesson traces to the expectation it serves and coverage gaps become visible at objective level.
- Teacher-owned. The teacher edits, approves and signs off, and the approval is recorded.
How it differs from adjacent products
| Approach | What the AI does | Who is accountable |
|---|---|---|
| AI tutoring | Interacts directly with the pupil | Ambiguous — the model mediates learning |
| Adaptive learning | Sequences content by prior performance | The algorithm sets the path |
| Chatbot lesson planning | Produces a document on request | Teacher, with nothing to check against |
| Pedagogy with AI | Drafts against a stated model and curriculum objective | The teacher, explicitly and on the record |
What it looks like in practice
1. Planning against the objective
A teacher selects a curriculum objective, a class and an instructional model. Edves drafts a lesson sequence with cognitive demand tagged per task, adaptation for the specific pupils in that class including those with SEND, and the check for understanding that will evidence the objective. The teacher edits it. What is taught is the teacher’s lesson.
2. Coverage that is real
Because plans carry objective references, a subject leader can see what was genuinely taught against the curriculum, and specifically which objectives appear in planning but never in assessment. That third category is where most of the gap between internal confidence and external results lives. See curriculum and teaching.
3. Adaptation rather than differentiation by outcome
Adaptation for pupils with SEND is generated from what the pupil’s plan actually specifies, not from a general category. This connects directly to Individual Support Plans and to the inclusion evaluation area.
4. Evidence for observation and appraisal
The planning record is the evidence an observer would otherwise assemble by hand, and it feeds the CPD trail. See teacher development and CPD.
Why the anchoring is the whole point
An unanchored model produces a plan that reads like a good lesson and is not one. Anchoring to a model and a curriculum objective turns a plausibility engine into something a subject leader can audit — and turns planning into the coverage evidence a school needs anyway.
It also protects the teacher. Under a framework that asks what was taught, to whom, and with what adaptation, a record showing which lesson addressed which objective and that the teacher approved it is a materially better position than a memory and a folder of slides.
Workload, honestly
The workload case for AI planning is real but frequently overstated. Drafting is faster; reviewing is not free. A teacher who accepts AI output uncritically saves time and produces worse lessons; a teacher who reviews properly saves less time than the marketing suggests and produces better ones.
The honest claim is that Edves shifts teacher time from producing a document to making pedagogical decisions about it. That is a better use of the hour, and it is what leadership and governance evidence on workload should be able to demonstrate — including what the school stopped doing.
Where AI is and is not used
Worth stating plainly, because staff will ask and DfE guidance expects schools to know:
- AI drafts planning, adaptation suggestions, feedback language and coaching prompts.
- AI does not assign assessment grades, make employment recommendations, decide SEND provision, or communicate with pupils unsupervised.
- Pupil data is not used to train general-purpose models.
- Every AI-drafted artefact carries a record of the human who approved it.
See DfE generative AI standards for how Edves is assessed against the published standards.