AI Marking Software and School Work Scrutiny
Work scrutiny - the book look, the work sample, the trawl through a class set to see what marking and feedback actually look like in practice - is one of the most common quality-assurance activities in any school. It is how leaders check that the marking policy is being followed, that feedback is timely and specific, and that standards are consistent across a department. As more teachers use AI to support their marking, scrutiny has to make sense of that too. For school leaders, it is worth thinking deliberately about where tools like AI marking software fit into work scrutiny: what to look for, how the tool changes what good practice looks like, and why, used well, it makes a scrutiny easier rather than harder. This guide is for headteachers, deputies, and middle leaders running work sampling.
What Work Scrutiny Is Actually Checking
A good book look has never been about counting ticks. The questions that matter are whether feedback is specific enough for a student to act on, whether it is consistent across a class and a cohort, whether it follows up - whether students have responded to it - and whether the standard applied to one student matches the standard applied to another. Those are questions about the quality and consistency of feedback, not its volume. The best scrutiny frameworks have been moving away from quantity for years, in line with the Department for Education's workload guidance, which warns against marking practices that generate effort without improving learning.
That focus matters when AI marking enters the picture, because it changes how some of those questions are answered while leaving the questions themselves intact. A scrutiny should still ask "is this feedback specific, consistent, and acted upon?" - it just needs to understand how a teacher using a tool arrived at it. Leaders who frame scrutiny around outcomes rather than mechanics will find AI marking fits naturally; those who frame it around the quantity of handwritten comments will find it confusing for the wrong reasons.
How AI Marking Changes What You See
When a teacher uses GradeOrbit, they define their marking criteria once and the tool marks every student's work against that same rubric, returning a suggested grade, a transcription, and categorised feedback the teacher then checks, edits, and confirms. The visible result in a class set is feedback that is unusually consistent: the same standard applied to the first script and the thirtieth, the same criteria referenced throughout - and typically more specific than a tired teacher produces by hand at the end of a long marking session.
For a scrutiny, that consistency is a strength to recognise, not a uniformity to be suspicious of. The crucial thing for leaders to understand is that the teacher remains the assessor of record: the AI produces a first pass, and the teacher checks and adjusts every grade and comment. Good practice with the tool looks like feedback that is criterion-referenced, consistent, and clearly engaged with by the teacher - not feedback that has been rubber-stamped without a read. A scrutiny conversation can ask a teacher how they used the tool and where they overrode it, which surfaces exactly the professional judgment a book look is meant to evidence.
Consistency as the Headline Benefit
The single hardest thing to achieve in a manual book look is consistency across a cohort and across markers. Drift is real: the same teacher marks differently when tired, and two teachers in a department mark the same work to different standards. A fixed, teacher-defined rubric applied to every script is the most direct answer to that problem, which is why AI marking tends to make a scrutiny more reassuring rather than less. Where a department has standardised on a shared rubric, the consistency a scrutiny is looking for is built in from the start.
This connects scrutiny to moderation, which asks the same question from a different angle. A department whose marking is consistent under scrutiny is a department whose moderation sample is likely to hold up.
What to Look For in a Scrutiny Where AI Is Used
For leaders running a book look where teachers use AI marking, a few practical pointers help. Look for feedback that is specific and criterion-referenced rather than generic - the tool supports this, but the teacher's check is what guarantees it. Look for evidence that students have acted on the feedback, which is unchanged by how the feedback was produced. Ask teachers to talk through a couple of examples: where did they agree with the tool, where did they override it, and why? A teacher who can answer that is demonstrably exercising judgment. The answer that should prompt a follow-up conversation is "I just used what it gave me" - not because the tool is wrong, but because rubber-stamping any first pass, human or AI, is the practice scrutiny is meant to catch.
Keeping It Supportive, Not Surveillance
Work scrutiny is most effective when teachers experience it as developmental rather than as inspection. The same is true of how leaders discuss AI marking within it. The honest framing is that the tool exists to reduce workload and improve the timeliness and consistency of feedback - and that any time it saves belongs to the teacher and to the higher-value work, not to an expectation of more marking. A scrutiny that treats AI marking as a support to be encouraged, with professional judgment kept at the centre, builds trust; one that treats it as something to be policed undermines both the tool and the relationship.
Privacy and Professional Responsibility
A point worth raising in any leadership conversation about marking tools is data handling, because teachers and parents will ask. For individual teacher use, GradeOrbit never stores uploaded student work - it is processed and discarded once results are returned, and teachers redact identifying details before upload, with the redaction burnt in on their own device. For schools adopting at an institutional level under a data processing agreement, the school journey persists work deliberately so staff can return to results, with defined retention windows. Either way, knowing the data model lets leaders answer questions confidently during and after a scrutiny.
Bringing It Into Your Scrutiny Cycle
For leaders planning the next round of work scrutiny, the practical step is to make sure your framework asks about outcomes - specificity, consistency, student response - rather than the mechanics of how marks were produced, and to treat AI marking as a legitimate, supported part of good practice rather than an anomaly. Update the questions your scrutiny asks so they make sense for teachers using a tool, and use the scrutiny conversation to surface the professional judgment the tool is designed to support. Our broader SLT guide to AI marking software sets this in the context of whole-school improvement planning.
Try GradeOrbit Ahead of Your Next Book Look
If your school wants marking that stands up to scrutiny - consistent, criterion-referenced, and clearly teacher-led - the most useful first step is to see the workflow yourself. GradeOrbit lets a teacher define their criteria, mark a real class set against them consistently, and keep full control of every grade, which is exactly the kind of practice a healthy work scrutiny is designed to find. New accounts get a small allocation of free credits to trial it on real work.
Visit gradeorbit.co.uk to learn more and get started, and to see how it could fit your school's approach to marking quality assurance.