Mark writing in minutes.
AI proposed,
teacher decided.
Graidus places a student's writing on ACARA's National Literacy Learning Progression v3, with verbatim evidence, the indicators met, and what to teach next. Every placement is a proposal your teacher accepts or overrides. Always.
Two teachers. The same piece of writing. Two different judgements.
Placing student writing against the progressions is slow, and it is inconsistent inside one staffroom, let alone across a network of schools. So it either doesn't happen often enough to show growth, or it happens and nobody entirely trusts the result.
A class set is an evening
Reading, transcribing, cross-checking indicators, writing it up. Done properly against five sub-elements, one round of samples eats a week of a teacher's own time.
The standard drifts
Two experienced teachers can place the same piece a level apart, and both be defensible. Multiply that across year levels and campuses and the data stops meaning anything.
The evidence disappears
Six months later, nobody can say which words in the writing led to the judgement, so the placement can't be explained to a parent, a new teacher, or a review.
Four steps from a photo of student work to a defensible placement.
No re-keying, no guesswork, no black box. Graidus does the heavy lifting and hands the judgement back to the professional in the room.
Capture the work
Drag in a photo, scan or PDF of handwritten or typed work, and multi-page is fine. Graidus transcribes it and keeps the raw original intact.
Place on the progression
Each of the five sub-elements is placed on NLLP v3, with a confidence score, verbatim evidence quotes, indicators met, and what comes next.
Accept or override
The teacher confirms the placement or overrides any sub-element with a reason. Every override is recorded. AI proposed, teacher decided.
Build the picture
The sample and its assessments are saved against the student, so growth is visible over time, not judged on a single piece.
One quiet workflow that does the parts teachers dread.
Student roster
Every student grouped Foundation to Year 6, searchable, with trend arrows for growth between samples. Import a CSV and bulk-promote a year at the end of term.
Progression timeline
A per-student history of every sample with colour-coded sub-element chips, a "what to teach next" panel paired with classroom activities, and side-by-side sample compare.
Cohort view
A class-level snapshot per year: distribution across Low / Basic / Sound / High, per-sub-element averages, and override rates, all mapped to your own reporting bands.
Parent-summary notes
Short, factual Strengths / Focus / Next-steps notes a teacher can lift straight into their own parent report, drafted from the actual evidence and never invented.
Print & PDF reports
A clean, printable per-student summary for files, meetings and handover. The same evidence, formatted for paper.
Ask the marker
A grounded assistant that answers from this student's real samples, cohort data and the rubric. Ask what to teach next, why an override was made, or for a planning prompt.
Not a vibe score. A judgement your school can stand behind.
Generic AI graders hand back a number and ask you to trust it. Graidus is built the other way around, for the realities of assessing a child's writing.
Published indicators, not a hidden scale
Every placement maps to observable NLLP v3 indicators. You can see exactly which were met and which come next, and so can anyone who questions it.
Evidence quoted verbatim
Evidence quotes are exact substrings of what the student actually wrote, so any placement traces straight back to the page. That matters most for spelling.
Override is first-class
Override isn't an escape hatch buried in a menu. Every assessment is "AI proposed, teacher decided", and that decision is logged with a reason.
Built for growth, not grades
One sample is a moment, not a verdict. Graidus tracks a student across samples and years, so the story is progression rather than a single snapshot.
The questions a principal asks, answered before they're asked.
- Privacy Act and the APPs, by designData handling is built around Australian privacy obligations and state department expectations, not bolted on afterwards.
- Aligned to the national GenAI frameworkMapped to the Australian Framework for Generative AI in Schools, so it fits the policy your sector already follows.
- Residency and retention you controlClear answers on where children's data lives and how long it's kept, with sensible defaults for a school context and no training on identifiable student work.
- Human accountability, recordedBecause a teacher accepts or overrides every placement, accountability stays with a professional, and the decision is logged for audit.
We're taking a small number of schools into Term 1, 2027.
Not a trial you're left alone with. The pilot runs for one term, on your own students' writing, and your literacy leader shapes what gets built.
What you get
- Everything, unmetered. All five sub-elements, every teacher, as many samples as you want to run.
- Set up for you. We import your roster, onboard your staff and run a session with your literacy leader in week one.
- A direct line. You talk to the person building it, and what you tell us changes the roadmap.
- Your rate, locked. Your price is agreed in writing before you start and held through 2028, so you'll never be quoted against later.
- The privacy pack. APP mapping, residency, retention and GenAI framework alignment, ready for your leadership team.
What we ask
- Actually use it. At least two rounds of samples across a year level or two. Enough to see growth, not just a demo.
- Half an hour a fortnight. With your literacy leader, so we hear what's wrong while it's still cheap to fix.
- A short letter of intent. Non-binding, at the founding rate, signed before the term starts.
- Permission to learn from it. The right to use de-identified samples and override decisions to improve the marking. Never identifiable, never shared.
- Tell people if it works. A reference call or a short quote at the end of the term, only if you'd actually recommend it.
Why now, and not next term. Australian schools set next year's budget in Term 3, and purchase orders issue between December and February. If Graidus isn't in front of your leadership by the end of September, the realistic start date is 2028, not 2027. That's the whole reason intake is open now.
Bring three of your students' writing samples.
Tell us a little about your school and we'll set up 20 minutes. We'll mark your samples live, in front of you, and you can tell us where we got it wrong.
If it isn't right for you, we'll say so. We'd rather have five schools using it properly than fifty who signed up and stopped.
What principals ask us first.
Is it available now?
Graidus is in build, and founding schools come on board for Term 1, 2027. If you want it in a classroom this term, we're not the right fit yet, and we'll tell you that rather than sell you a waitlist.
Which year levels and which parts of the progression?
Foundation to Year 6, across the five writing sub-elements of NLLP v3: Creating texts, Grammar, Punctuation, Spelling, and Handwriting and keyboarding. Years 7–10 and the numeracy progressions are on the roadmap, not in the product today.
Does it handle handwritten work?
Yes, that's the normal case. Photograph or scan the page, including multi-page pieces, and Graidus transcribes it while keeping the original image attached to the sample. Teachers can correct the transcription before marking, and the evidence quotes always point back to the real words on the page.
Does this replace teacher judgement?
No, and we've built it so it can't. There is no mode that saves a placement without a teacher accepting it, and there won't be one, not even if a customer asks. The AI proposes, the teacher decides, and the decision is what gets recorded.
The professional judgement of the teacher is the point of the product. The reading, the cross-checking and the write-up are what we're taking away.
Where does student data live, and is it used to train AI?
Children's data stays in Australia. Identifiable student work is never used to train models. Founding schools are asked separately, in writing, for permission to use de-identified samples to improve the marking. You can say no and still be a founding school.
The full privacy pack, covering APP mapping, residency, retention and GenAI framework alignment, is a document we'll send you rather than a conversation you have to chase.
What does it cost?
An annual whole-school licence, banded by enrolment, aligned to the school year and invoiced on a purchase order. Unlimited teachers, unlimited samples. We don't meter marking, because a tool teachers ration is a tool that never becomes a habit.
Ask and we'll send the rate card the same day. Founding schools get their rate locked before they start.
We already use Brightpath, PAT or Essential Assessment. Why this as well?
Those tools put students on a calibrated scale, usually from a purpose-run assessment task. Graidus does something narrower and more everyday: it places the writing your teachers already collect against the national progressions, shows the evidence for each placement, and keeps the record over time.
Plenty of schools will run both. If what you actually need is a standardised scale for reporting, say so and we'll tell you honestly whether we're worth your time.
Can a diocese or school group roll this out across schools?
Yes, and that's where it gets interesting. Consistency of teacher judgement is hard inside one school and much harder across eighty. Cross-school moderation reporting is being built for exactly that conversation. Get in touch and tell us what your system leaders would need to see.