The question no staffroom has fully settled
Picture a Grantham teacher marking homework on a Tuesday evening. A pupil has handed in a well-structured essay plan — clear argument, logical progression, decent vocabulary. The teacher suspects, but cannot prove, that it started life as a ChatGPT prompt. Nothing has been copied and pasted. The pupil has clearly read it, adjusted it, and added their own examples. So: was that help, or was it doing the work for them?
That question is sitting, largely unanswered, in staffrooms across South Kesteven right now. It sounds simple — almost procedural — but pull at it and it opens onto something much larger: what education is actually supposed to do. Is the point to produce the essay plan, or to develop the mind that could produce one? And if the answer is the latter, does it matter which bits a pupil did alone? Schools in Grantham are working through exactly this tension, without a settled answer, and without much precedent to draw on.
What Grantham schools are actually doing
Four Grantham-area schools offer a useful cross-section of where practice currently sits — and the differences between them are as instructive as any single approach.
At The King's School, the most visible AI investment is structural. The school uses the CENTURY platform, an adaptive-learning system that adjusts content to each pupil in real time and flags specific knowledge gaps to teachers. The technology is doing something a single teacher with thirty students cannot easily replicate: tracking individual progress at a granular level and responding to it.
Kesteven and Grantham Girls' School has taken a different route. Rather than deploying an AI tool for instruction, KGGS has woven AI into the curriculum as a subject of scrutiny. Pupils use the BBC Bitesize guide to AI during tutor sessions and explore questions of AI ethics through structured debate. The emphasis is on critical interrogation of the technology — asking what it can and cannot do — rather than fluency with it.
Grantham Prep School is working on a related but distinct instinct: scepticism as a taught skill. Pupils are trained explicitly to verify AI-generated answers rather than accept the first result they see. The school is treating the habit of checking as something that needs deliberate practice, not something students will naturally develop on their own.
The most formal governance step in the district came from Westminster School in South Kesteven, which published a written AI policy in March 2025. The document covers staff, governors and pupils, and names generative tools including ChatGPT and Google Bard by name — making it an unusually concrete piece of institutional decision-making at the school level.
Taken together, the picture is cautious and deliberately bounded. AI is being introduced into computing lessons, STEM workshops and specific curriculum slots — not rolled out across every classroom at once. Grantham College's Institute of Technology, with its digital design suites and engineering robotics facilities, signals where some of that trajectory is heading at post-16 level, but for now the dominant local posture is opt-in, selective, and institution-led.
What national policy says — and admits it doesn't know
Central government has been moving quickly, by its own standards. The Department for Education published its first formal AI guidance for English schools in June 2025 — updated again in August — and the framework it sets out is clearer on some questions than others.
For teachers, the guidance is broadly permissive. Staff may use AI across lesson planning, resource creation, marking, feedback and administration. The one firm line is accountability: professional responsibility for every output remains with the teacher, not the tool. Alongside this, the DfE is funding a set of subject-specific AI applications — systems designed to assess handwritten work, provide feedback on hand-drawn geography maps, and identify errors in student soldering on circuit boards. These are practical, bounded uses rather than speculative futures.
The TechFirst programme extends the ambition to pupils directly, targeting around one million secondary students each year with hands-on technology and AI exposure — a signal of how seriously government is taking AI literacy at scale.
But the guidance contains an admission that is more revealing than the commitments. The DfE states plainly that evidence on the benefits and risks of pupils themselves using generative AI is still emerging. This is not a gap that researchers or critics have uncovered — the government has named it in its own published document.
That matters for how Grantham schools' caution should be read. An opt-in, selective approach is not a sign of being slow to adapt; it is a reasonable response to a genuine evidence vacuum that even national policy has not yet filled.
The deeper shift: from recalling facts to doing something with them
Behind the staffroom uncertainty about essay plans sits a sharper question than it first appears. The Tony Blair Institute's September 2025 analysis of AI in English schools argues that the technology does not simply change how students learn — it changes what learning is supposed to produce. In a world where a language model can retrieve and recombine information on demand, the measure of educational success shifts away from knowledge-recall and toward what students can do with knowledge: depth of thought, original argument, collaborative reasoning.
That reframe makes the Tuesday-afternoon essay-plan dilemma something more than a marking headache. The risk the Institute names is 'cognitive offloading' — students using AI to skip the uncomfortable work of forming an idea rather than to extend it. The discomfort of not knowing quite how to begin an essay is not a flaw in the task; it may be the task. If AI removes that friction, it can remove the thinking along with it.
The mechanism this identifies is what makes the question 'did the pupil use AI?' less useful than 'did the pupil think?' A student who uses an essay-plan suggestion as a scaffold to argue against it has done something different from one who copies the structure and fills in the blanks — yet both have 'used AI'.
It is important to be clear about what this analysis is and is not. No attainment data from Grantham schools — or from English schools more broadly — yet exists to confirm that cognitive offloading is happening at scale, or to measure its effect. The Tony Blair Institute's framing is an analytical argument, not a measured outcome.
What responsible adoption actually requires
The tools that Grantham schools have chosen — a CENTURY adaptive platform here, a BBC guide in tutor time there — matter less than a prior question those choices force into view. What structures need to be in place before a pupil opens a chatbot?
The Tony Blair Institute's analysis of AI-ready education systems points to Estonia as instructive not for its specific software but for its sequencing: teacher training and national governance were established as prerequisites before AI tools were introduced at scale, rather than added retrospectively when problems emerged. England's DfE guidance, published in June 2025, places professional responsibility firmly on teachers — but has not yet mandated equivalent training requirements to support that responsibility.
What the South Kesteven evidence suggests is that some schools are already working through this logic at institutional level. Westminster School's written AI policy — already described — is a governance decision first: it draws the perimeter before pupils reach the tools. KGGS's approach does something adjacent but distinct: embedding ethics questions into the timetable treats governance as curricular rather than merely procedural. The distinction matters. A policy document sets rules; a debate session builds the capacity to question.
Whether either approach improves outcomes is genuinely unknown. No attainment data yet links these structural choices to measurable gains — for Grantham schools or English schools more broadly. What they are building is the architecture that would make outcomes trackable: not proof that responsible adoption works, but the conditions under which you could begin to find out.
What the question means for students growing up in Grantham now
Students sitting in Grantham classrooms in 2026 are being shaped by choices made in a window before anyone has the data to confirm those choices were right. That is not a criticism of local schools — it is the honest condition for every school in England right now.
What distinguishes the local picture is less about which platforms are running than the orientation being built alongside them. A pupil at KGGS who has spent tutor time questioning AI outputs and arguing about what the tool should and should not do leaves that session with something a pupil at a school with no policy does not: a trained reflex to pause before accepting. Whether that reflex produces measurable gains in attainment, no one can say yet. But it is the kind of habit that precedes understanding rather than substituting for it.
The question the TEDx Grantham framing keeps returning to — what counts as thinking when AI does the work? — has no settled answer. The most honest observation from the local evidence is that some Grantham schools are at least teaching students to hold that question open. In a period where no one — not researchers, not policymakers, not the schools themselves — fully knows what AI-assisted learning produces, that orientation may be the thing that matters most.