The Department for Education (DfE) recently issued new guidance encouraging the use of artificial intelligence (AI) to reduce teacher workload. As reported by the BBC, AI can help automate routine tasks such as marking and writing letters and resources. It’s a step in the right direction but also one that demands precision, transparency, and, above all, trust.
At Graide, we believe that AI can and should play a meaningful role in reshaping how assessment works. However, not all forms of AI are equal. And that matters especially in education.
The Problem with Generative AI in Marking
The DfE's guidance is careful and clear: any use of AI in marking must be reviewed by teachers. Why? Mainstream generative AI tools such as ChatGPT have the potential to "hallucinate", meaning they can generate responses that are confident yet inaccurate or completely fabricated. As Emma Darcy pointed out in the BBC article, these tools can invent quotes, facts, or misinterpret student work.
Many popular AI tools today rely on generative AI, and that’s a serious risk in an educational setting. Inaccurate feedback can confuse students, undermine trust in the system, and even widen achievement gaps if unchecked.
Graide’s AI Is Different: No Hallucinations, Just Accuracy
Graide’s AI is not generative, not LLM-based. It doesn’t make things up. Instead, it’s designed specifically for assessment in education. Consider it as an automated process that is structured and grounded in evidence. It speeds up marking by identifying patterns, applying consistent rules, and using rubrics to assess responses in ways that teachers can review and control.
Teachers maintain control at all times. They can edit, approve, or override the AI’s suggestions, giving them more time to focus on pedagogical decisions, not repetitive admin.
So How Does Graide Work?
- Aligned to rubrics: Graide uses pre-set mark schemes and models to ensure marking is criteria-driven.
- Transparent and traceable: every AI-generated comment is linked to student evidence, ensuring teachers can verify it instantly.
- No generative text: Graide does not create freeform responses. It classifies and matches student responses based on rigorously tested logic.
- Teacher-in-the-loop: Our model was built to assist, not replace. Teachers can adjust feedback and scoring in seconds.
The result? Accurate, scalable, and fair assessment without the fear of “hallucinated” feedback.
Supporting the DfE’s Vision: Safely
We fully support the DfE’s push for AI to “cut workloads” and “free teachers from paperwork”. And we echo the ASCL’s call for more investment so schools can access trustworthy tools.
As ASCL’s general secretary, Pepe Di’Iasio, noted, the promise of AI is real, but so are the barriers. Budgets are tight, and schools need solutions that are reliable and ready today, not tools that introduce risk.
We also understand the need for policy clarity, especially around transparency. Because Graide doesn’t generate creative content but rather automates structured assessment processes, the reporting burden is far lower and the output far easier to trust.
The future of education will involve AI. That’s not up for debate. But the kind of AI we use matters.
Tools like Graide represent a new category of education-specific AI: precise, auditable, and designed to work with teachers, not replace them.
If we want to embrace AI without compromising trust, accuracy, or professional standards, we need to champion platforms that were built for education from the ground up.
And that’s exactly what Graide is.
Want to see how Graide uses AI to reduce marking workload without the risks of hallucination? Book a clarity call to learn more.
.png)