The decision landing on every local teacher's desk
Somewhere in Grantham this week, a teacher opened a ChatGPT window before the school day started. Perhaps they were drafting a SEND support note, roughing out a lesson plan, or trying to get a parent letter done before registration. Nobody told them to. Nobody told them not to. There is no policy on the wall.
This is not a future scenario. Across the mix of primaries, secondaries, and trust-linked schools that serve South Kesteven, AI tools are already woven into professional routines — often quietly, often without any formal framework governing their use. A Twinkl survey of more than 6,500 UK teachers found 60% using AI professionally by January 2025, double the proportion from two years earlier. Primary adoption stood at 53.3%.
The problem is that the Department for Education has set a deliberately permissive framework: schools may choose their own rules, provided they stay within data-protection, child-safety, and intellectual-property law. That leaves headteachers and classroom teachers in Grantham making calls that feel, as one national union survey put it, well above the level of guidance they have been given — balancing the clear pull of workload relief against real and unresolved questions about whose data is entering which system.
What teachers are actually using AI for
The clearest measure of what teachers are actually doing comes from a randomised controlled trial commissioned by the Education Endowment Foundation and conducted by NFER. Across 259 teachers in 68 schools, those using ChatGPT for lesson preparation saved roughly 25 minutes per working week — a 31% reduction in planning time. That is not a dramatic transformation, but for a profession routinely citing workload as its primary stress driver, it is a consistent and reproducible gain. The 64% of frequent AI users in the wider Twinkl survey who reported saving one to five hours weekly, and the 67.8% who said AI had reduced their workplace stress, point in the same direction.
The most common applications are administrative and preparatory: generating classroom resources, drafting lesson plans, writing end-of-term reports, and creating differentiated materials for pupils with SEND. That last use has now received an explicit policy signal. Updated DfE-backed guidance published in May 2026 states that schools can use AI to draft SEND support plans and staff guides for pupils with special educational needs, provided school leaders retain oversight and professional responsibility for every output. Coming from a department that has largely left schools to set their own course, this is a notable shift in official tone — from permitting professional AI use to actively suggesting specific applications.
The distinction is worth holding onto: these are tools teachers use for their own planning and paperwork, not AI systems deployed directly with pupils in lessons. That boundary — between professional productivity and pupil-facing instruction — is where the safeguarding questions become considerably sharper.
How Lincolnshire schools are drawing their own lines
Three Lincolnshire schools offer the clearest regional picture of what bounded AI adoption looks like in practice.
The King's School Grantham is understood to use CENTURY Tech — an AI-driven platform that draws on learning-gap data and neuroscience-informed intervention to identify where pupils are struggling and target misconceptions before they become entrenched. That sits alongside Microsoft 365 and Teams for day-to-day cloud working. The combination is typical of how secondary schools in the area are approaching EdTech: tools embedded in existing workflows rather than deployed as standalone experiments.
Rosebrook Primary, in Lincolnshire, provides the most detailed local school AI policy in the public domain. Staff use ChatGPT to reduce planning and administrative workload; senior leaders use AI tools for leadership administration. The school conducts regular risk assessments and enforces strict GDPR compliance. Its most deliberate editorial choice is a clear ceiling: AI will play no part in any decision about a pupil's future — not career guidance, not subject choices, not progression recommendations. That line is a policy decision, not an absence of policy, and it reflects exactly the kind of institutional judgement the DfE framework leaves to individual schools.
Boston's Tower Road Academy joined just four other schools nationally in piloting Pocketalk AI translation — devices covering 84-plus languages — to help EAL pupils integrate into school life. The pilot is a useful example of a targeted, narrow deployment: one tool, one identified need, measurable potential impact.
The pattern across all three is consistent. These schools are not waiting for national guidance to catch up; they are drawing their own lines, setting deliberate limits on what AI can decide, and treating adoption as a managed risk rather than an open invitation.
A permissive framework that puts school leaders on the spot
The DfE's current position is not ambiguity by accident. Its published guidance allows schools to set their own AI rules within three existing legal frameworks — data protection law, Keeping Children Safe in Education (KCSIE), and intellectual property rules — but does not require them to do anything specific. Teachers may use AI; they must check its outputs; professional responsibility stays with them. Beyond that, the department has deliberately left the field open.
The consequence shows up plainly in two consecutive NEU State of Education surveys. In both 2025 and 2026, 66% of teachers reported working in schools with no AI policy specific to student use. Nearly half — 49% — were at schools with no AI policy at all, for staff or pupils. These are not figures from a moment of initial confusion; they describe the situation two years into widespread adoption.
KCSIE 2025 added AI to the existing safeguarding framework in a meaningful but limited way. Schools are now expected to complete a Data Protection Impact Assessment before deploying any AI EdTech tool, and to evidence a 'balanced approach' rather than a blanket ban. What it does not provide is a policy template, a curriculum framework, or a clear standard for what safe pupil-facing AI use actually looks like in practice.
Two DfE initiatives — the EdTech Evidence Board pilot launched in January 2025 and the EdTech Impact Testbed programme launched in May 2025 — are designed to build that evidence base. Neither has yet produced binding guidance. Until they do, the practical effect for a headteacher in Grantham or anywhere else in South Kesteven is the same: write your own policy, make your own calls, and carry the professional and legal responsibility that goes with it.
Early years: the corner where the debate gets harder
Reception-age children complicate the debate in ways the secondary-focused national conversation rarely addresses.
The legal baseline is straightforward: children in early years and reception cannot legally consent to data processing. Any AI tool that profiles, assesses, or tracks a child under five requires explicit parental consent and, under KCSIE 2025, a Data Protection Impact Assessment before deployment. That requirement applies whether the tool is a sophisticated adaptive platform or a simpler planning assistant that holds pupil data.
Beyond data law, there is a developmental question that evidence cannot yet fully answer. AI assessment tools generate plausible-sounding progress reports — but a four-year-old's development is non-linear in ways that make algorithmic attainment summaries particularly susceptible to error. Using AI to draft a phonics lesson plan is a categorically different act from using it to generate a developmental assessment of a specific child. The first is a staff-facing productivity tool; the second produces a record that may influence decisions about that child's support needs.
The statutory EYFS framework is also structurally ill-matched to algorithmic planning: its emphasis on play-based, child-led learning depends on practitioner observation and in-the-moment responsiveness that adaptive content tools are not designed to support.
Neither the DfE nor Ofsted has published EYFS-specific AI guidance. That is a genuine gap rather than a scandal — early years simply has not been the focus of the national AI-in-schools debate — but it means primary leaders in Grantham who run nursery or reception provision are navigating this corner without much to draw on.
What Grantham school leaders need that policy does not yet give them
School leaders in Grantham are not asking for permission to use AI — most of their staff already are. What they need is something national policy has not yet produced: practical templates built for primary and EYFS settings rather than secondary schools, a digestible version of the best available trial evidence, and shared legal infrastructure for DPIA compliance that individual primary schools cannot realistically build alone.
The NFER randomised controlled trial — 259 teachers across 68 schools — is the most rigorous evidence available on what AI actually does in classrooms. But awareness of it among headteachers outside research networks remains patchy. A school leader making a policy call in 2025 had to find that evidence independently. Most did not have the time.
DPIA requirements under KCSIE 2025 place a genuine legal burden on schools without in-house data protection expertise — which describes most Lincolnshire primaries. Running a separate assessment for every new EdTech tool is not a realistic expectation without external support. The schools managing this well — Rosebrook's explicit ceiling on pupil-future decisions, Tower Road's targeted EAL pilot — did so through deliberate senior leadership investment. That model is not easily replicated across a district without trust-level or local-authority coordination.
Whether Lincolnshire County Council or the multi-academy trusts operating across South Kesteven will step into that gap remains an open question. The national framework has been set deliberately loose, which is a policy choice, not an oversight. The decisions filling it are being made in staffrooms and SLT meetings across Grantham right now — and they will shape local practice long after the guidance catches up.