Feedback is where AI can genuinely reduce workload without reducing quality, provided it's used to support your judgement rather than replace it.

Practical uses

  • Generating a bank of specific, actionable feedback phrases for common error types in a subject, which you select from rather than writing from scratch each time
  • Turning a quick verbal note into a written comment structure
  • Drafting whole-class feedback summaries after marking a set of books, based on the patterns you've noticed

What not to automate

Don't feed pupil work directly into a public AI tool to generate feedback about that specific piece of work — this risks both data protection (if it's identifiable) and quality (AI hasn't seen the lesson, the pupil's starting point, or your success criteria). Use AI to build your toolkit of phrasing, not to mark for you.

Whole-class feedback, done well

One of the best-value uses here is whole-class feedback: after marking a set of books yourself, describe the patterns you noticed ("half the class struggled with apostrophes in contractions, a few used excellent varied sentence openers") and ask AI to structure this into a clear whole-class feedback sheet. You did the marking and the noticing — AI just saves you the writing-up time.

Worth knowing: AskColin's own do/don't guidance lists "drafting generic feedback comments for review" explicitly as an appropriate use — always with the requirement that a member of staff reviews it before use. See the full guidance →

Key takeaways

  • Use AI to build a phrasing toolkit, not to mark pupil work directly.
  • Whole-class feedback write-ups are a strong, low-risk use case.
  • Never feed identifiable pupil work into a public AI tool for feedback.