Teaching in New Zealand has always demanded a lot. But in recent years, the workload has reached a tipping point that the profession can't ignore. The 2024 PPTA (Post Primary Teachers' Association) workload survey found that secondary teachers are routinely working 50+ hours per week, with marking consistently ranking as the single biggest contributor to unsustainable workloads.
AI essay marking isn't a silver bullet. But for many New Zealand teachers, it's becoming the difference between a sustainable workload and one that drives people out of the profession.
This piece is for teachers who are curious about AI marking but haven't taken the leap yet. We'll cover the reality of the technology, the legitimate concerns, NZQA's position, and a practical starting point.
📋 Table of Contents
The NZ Teacher Workload Reality
If you teach an essay-based subject to 90+ students across multiple classes, you know the arithmetic. Three classes of 30 students, each submitting an essay of 600–1000 words every two to three weeks. Even at the optimistic end of 7 minutes per essay, that's 10.5 hours per round of marking — before you've read a single research article, prepared a lesson, or attended a meeting.
This is one of the primary reasons New Zealand is experiencing a teacher retention crisis. It's not just the pay (though that matters). It's the volume of invisible work that never ends. And for many teachers, essays represent the most time-consuming part of that work.
AI marking tools don't eliminate the work. But they restructure it in a way that many teachers find significantly more sustainable — from generating feedback from scratch (slow, draining) to reviewing AI-generated feedback and applying professional judgment (faster, more satisfying).
How AI Essay Marking Actually Works
There's a lot of misunderstanding about what AI marking tools do. They're not simply autocorrecting grammar, and they're not replacing teacher judgment. Here's the accurate picture:
- You provide the rubric. You give the AI your marking criteria — in NCEA terms, the Performance Descriptors for each grade level. This is critical: the AI doesn't invent its own standards; it applies yours.
- The AI reads each essay and generates an assessment. It identifies how well the essay meets each criterion, flags evidence from the text that supports its assessment, and produces a provisional grade.
- You review the AI's work. You look at the grade, the feedback, and the evidence. For clear cases (comfortably Achieved or clearly Not Achieved), this takes 2–3 minutes. For borderline cases, you spend as long as you need.
- You make the final decision. Nothing goes to students without your approval. You can override any grade, add personal notes, or revise the AI's feedback before releasing it.
The result: instead of generating feedback from scratch, you're reviewing and editing AI-generated feedback. For experienced teachers, this is consistently faster — and often more consistent, because you're applying the same criteria to every essay rather than recalibrating after each one.
AI Marking and the NCEA Framework
The NCEA's standards-based approach actually aligns well with AI marking. Because the criteria are explicit and well-defined in each Achievement Standard, you have clear rubric language to give the AI. The N/A/M/E scale has specific Performance Descriptors — the AI doesn't have to interpret abstract standards, it applies your criteria.
| Without AI Marking | With AI Marking |
|---|---|
| Read essay cold, form initial impression | Review AI's initial assessment with evidence |
| Re-read to confirm or revise judgment | Spot-check 2–3 passages to verify AI reasoning |
| Write feedback from scratch (5–15 min) | Review and personalise AI feedback (1–3 min) |
| Record grade | Approve and release (automated) |
| Total: ~8–12 min per essay | Total: ~2–4 min per essay |
For a class of 28 students, that's the difference between 3–4 hours and 60–90 minutes. Across a term with four major assessment rounds, that's 6–10 hours returned to you.
To explore how this works specifically for NCEA, visit the NCEA essay grading page and the GradingPen New Zealand page.
See How AI Marking Works with Your NCEA Standards
Upload your next batch of NCEA essays and see what AI-assisted marking looks like in practice. 15 free grades — no credit card required.
Start Free Trial →NZQA's Position on AI Tools
NZQA has been measured and clear in its position: teachers are responsible for all assessment decisions. AI tools can support the marking process, but the final grade is always a teacher judgment call.
This matters because it draws a clear line: AI as a marking assistant (supported) versus AI as the assessor (not supported). GradingPen is designed to operate entirely in the first category — it produces a provisional assessment, you make the final call.
NZQA has also been clear about the academic integrity side: the concern is students using AI to write essays, not teachers using AI to mark them. These are entirely different issues, and conflating them is a mistake that causes unnecessary anxiety about legitimate tools.
For the latest NZQA guidance on digital tools in assessment, always check the NZQA website directly — their guidance is updated regularly as the technology and policy landscape evolves.
Addressing the Legitimate Concerns
Concern 1: What if the AI gets it wrong?
It will, occasionally. That's why teacher review is built into the workflow — you're not rubber-stamping AI grades, you're reviewing them. When the AI is uncertain (often flagged automatically), you spend more time. When the essay is a clear case, you spend less. Your professional judgment applies where it matters most.
In practice, teachers using AI marking tools report about 85–90% initial agreement between AI grades and their own assessment. The disagreements are almost always on borderline essays — exactly where a second perspective is most valuable.
Concern 2: Academic integrity — are students using AI to write essays?
This is a genuine concern, but it's separate from the question of whether teachers should use AI to mark essays. AI writing detection has improved significantly, and many of the essays that look AI-written are actually just essays where students have structured their writing more formally (which is, ironically, what good essay feedback encourages).
Use your professional judgment on suspected AI-generated essays, and have a conversation policy ready. But don't let this concern prevent you from using a marking tool that could genuinely change your workload.
Concern 3: Student data privacy
This is the most important technical concern. Any tool handling student essays needs to meet privacy standards. Look for tools that: don't use student essays to train AI models, store data with encryption, offer a Data Processing Agreement for school-wide use, and allow data deletion on request. GradingPen provides all of these. Review the privacy policy carefully before school-wide deployment.
What to Tell Parents
If parents ask about AI marking, here's a straightforward way to explain it:
"I use an AI tool to help me provide faster, more consistent feedback on your child's essays. The AI gives me an initial assessment, which I then review and personalise. The final grade and feedback are always my professional decision — AI helps me work more efficiently, not differently."
Most parents will find this reassuring. The analogy to spell-check or a calculator is useful: those tools also help teachers work more efficiently without replacing professional judgment, and nobody questions them.
If a parent is specifically concerned about their child's data, direct them to your school's privacy policy and the tool's data handling documentation. Having that information ready in advance is good practice.
Getting Started Safely
The lowest-risk way to trial AI marking is with formative or practice work — essays that don't count toward a student's final grade. This gives you time to calibrate the tool with your own assessment standards before using it for summative marking.
- Create a free account at GradingPen's New Zealand page
- Select a recent NCEA standard you've marked by hand
- Set up your rubric using the Performance Descriptors from that standard
- Upload 5–10 essays you've already marked
- Compare the AI grades to your own — calibrate where needed
- Scale up to a full class on your next formative task
Most teachers find that after two or three practice rounds, they trust the tool enough to use it on summative work — with the same careful review they'd apply to any marking decision.