Formative feedback at scale
Draft next-step comments on extended writing so every pupil hears back sooner and more often, with the teacher reviewing and owning what is sent.
— AI use case · Education —
AI for marking
and feedback.
AI can give teachers time back and pupils faster feedback, and it can go badly wrong if it is deployed without validation, fairness testing, and a human owning every grade. Independent guidance from the Institute of AI, with no marking software to sell.
— State of play —
The workload is real,
so is the risk.
Marking is the largest single call on a teacher's time, and feedback is where most of the learning gain actually happens. Generative AI can draft formative comments, summarise where a cohort is struggling, and give pupils a response in minutes rather than a fortnight. Used well, it gives teachers time back and pupils faster feedback.
It is also one of the highest-stakes places to put AI to work. A marking or feedback tool touches children's data, shapes attainment records, and can influence grades and life chances. The same properties that make it useful, scale and speed, are exactly what make an unvalidated or biased system dangerous.
The Institute of AI is an independent professional body with no marking software to sell. We help education leaders separate the real benefit from the sales pitch, and adopt marking and feedback tools that governors, examiners, and regulators can stand behind.
Four
levels of professional accreditation, Student to Fellow, against IoAI's published competency framework
Five
pillars in the IoAI organisation maturity assessment: strategy, governance, skills, implementation, impact
Free
to sign the UK AI Readiness Charter, free for any school or trust
— Use cases —
Where AI helps with
marking and feedback.
The strongest cases keep a teacher in the loop and give time back to people, not platforms.
First-pass structured marking
Score against a rubric or mark scheme as a starting point, flagging borderline scripts for a human to confirm rather than replacing the marker.
Cohort gap analysis
Summarise recurring errors and misconceptions across a class or year group so planning and reteaching target the right things.
Report-writing support
Turn a teacher's bullet points and marks into consistent, personalised report prose, keeping the professional judgement with the teacher.
Self and peer assessment
Give pupils rubric-aligned prompts and worked exemplars so they can improve a draft before it reaches the teacher for final marking.
Accessible, differentiated feedback
Adjust reading level, offer audio feedback, and draft translations for EAL families for a bilingual colleague to check before sending, so comments land with the pupil they are written for.
— Governance —
What must be
governed.
Marking and feedback is high stakes. Every deployment we support has to answer each of these before it touches a pupil's grade.
01
Model risk and validation
Prove it marks accurately before it marks anything
A tool that scores work must be validated against human markers on your own scripts and mark schemes, not a vendor demo. Agreement rates, drift over time, and behaviour on unusual answers all need measuring before a single grade is affected.
100%
of scoring tools benchmarked against human markers first
02
Fairness and bias
Marking must not disadvantage any group
Language models can penalise EAL pupils, non-standard dialects, or particular writing styles. Feedback and scores need testing for disparate impact across groups, and a plan for what happens when a gap is found.
4+
pupil groups checked for disparate impact
03
Explainability
A mark a teacher cannot explain is not usable
Pupils, parents, and appeals panels are entitled to know why a piece of work received the feedback it did. Marking tools have to surface the reasoning against the rubric, not just a number, so a teacher can defend or override it.
0
black-box scores accepted without a rationale
04
Data protection
Children's work is personal data
Pupil scripts, attainment, and SEND context are sensitive personal data under UK GDPR. Any marking tool needs a DPIA, a lawful basis, clear retention, and assurance that pupil work is not used to train a vendor's model.
1
DPIA completed before go-live, every time
05
Human accountability
A person owns every grade
AI drafts, a teacher decides. For a regulated qualification, Ofqual's conditions and JCQ's instructions mean AI must not determine the mark at all: a named marker reviews the output and remains accountable for it, with the authority and time to override.
1
accountable human sign-off on graded work
06
Monitoring after go-live
Accuracy on day one is not accuracy on day 200
Models drift, cohorts change, and mark schemes evolve. Live deployments need sampling, agreement tracking, and a route for teachers to report bad feedback, so quality is watched rather than assumed.
Ongoing
sampling of AI-marked work for human review
07
Safeguarding
Pupil work can contain a disclosure
Extended writing marked at scale will sometimes contain a safeguarding disclosure. A marking tool needs a clear route to the Designated Safeguarding Lead, and staff who know that Keeping Children Safe in Education duties still apply when the first reader is a model.
DSL
a defined route to the safeguarding lead
— How the Institute of AI helps —
One professional body,
three kinds of help.
Advice, engineering, and professional standards from a single independent body, with no marking software to sell.
Consultancy
Independent advice
Standards-led advisory for heads, trust executives, and exams officers deciding whether and how to use AI for marking and feedback, with no marking software to sell.
- Tool evaluation and validation design
- Marking and feedback policy
- DPIA and governor briefings
Solution design
Systems built and handed over
Where an off-the-shelf tool will not do, the Institute of AI's engineering practice builds a marking or feedback system around your rubrics and data protection duties, then hands it to your team.
- Rubric-aligned feedback tooling
- Deployment on your existing systems
- Knowledge transfer to your staff
Professional body
Standards and the Charter
As the UK's professional body for AI, IoAI accredits organisations against a five-pillar maturity model, accredits AI professionals against a published competency framework, and maintains the UK AI Readiness Charter so your marking practice meets a recognised bar.
- Organisation accreditation
- Professional accreditation and CPD
- The UK AI Readiness Charter, free to sign
Put AI to work on marking,
properly.
A short call with the Institute of AI. Plain answers on where AI marking helps, where it does not, and how to deploy it safely. No marking software to sell.
— Go deeper
— Sector
AI in education
Our full programme for schools, trusts, colleges, and universities adopting AI with care.
— Consultancy
AI consultancy
Independent, vendor-free advisory for boards and leaders setting an AI position they can defend.
— Standards
The Charter
The UK AI Readiness Charter: five plain pledges that get your organisation ready for AI.