Edvolve

The article is intended for Headteachers · Teachers · School Leadership Teams

Ask most UK teachers right now whether AI has changed their working week and the honest answer is almost certainly yes. The National Education Union's most recent survey of its members, published in spring 2026, found that 76% of teachers are now using AI tools for day-to-day work, this is up from 53% just a year earlier. Most of that use is in resource creation, lesson planning and administrative tasks. A trial run by the Education Endowment Foundation and NFER found that structured use of tools like ChatGPT cut lesson planning time by around 31%, saving roughly 25 minutes per lesson without a measurable drop in resource quality.

Those are genuinely significant numbers, in a profession where the Department for Education's own workload survey has long shown teachers working close to 50 hours a week, with over a quarter of that time going to planning, marking and admin rather than pupils. So the case for AI easing teacher workload isn't speculative, it's already happening, at scale, in classrooms across the country.

But the same survey that reported that 76% adoption figure also reported something else worth sitting with - 66% of secondary teachers said they'd personally observed a decline in pupils' critical thinking as a result of AI use and a third of those teachers said they strongly agreed. That's not a fringe concern from a handful of sceptical staff rooms, that's two-thirds of secondary teachers, describing something they're watching happen in their own classrooms.

This article isn't going to land on a tidy conclusion that AI is simply good or simply bad for schools, we don't think that's an honest position and we don't think the evidence supports it either. What we want to do instead is look honestly at both sides because the schools navigating this well right now are the ones doing exactly that.

The case for AI easing the load is real

It's worth being specific about where AI is genuinely helping, because the benefits aren't vague or aspirational, they're showing up in how teachers actually spend their time. Administrative tasks are the most obvious win by drafting routine parent communications, summarising data, generating first drafts of reports, these are exactly the kind of repetitive, low-judgement tasks that eat hours without meaningfully improving outcomes for pupils. Time saved here is time that can genuinely go back into teaching, planning, or simply going home on time.

Lesson resource creation is the next biggest category, and the EEF/NFER trial data backs up what many teachers report anecdotally - a well-structured prompt can produce a usable first draft of a worksheet, quiz or starter activity considerably faster than building one from scratch.

Differentiation at scale is an area where AI has a less-discussed but real benefit. Producing three versions of the same resource pitched at different reading levels is exactly the kind of task that often gets skipped under time pressure, not because teachers don't want to differentiate, but because there genuinely isn't time to do it for every lesson, every day, AI tools have made this more achievable.

For a profession where workload pressure is consistently linked to staff retention and wellbeing, 84% of teachers reported a negative impact on their mental health from workload in one widely cited survey.

What the data also shows us and why it should give us pause

Here's where it gets more complicated and where we think the conversation often becomes too simplistic in either direction.

The same body of evidence pointing to time savings is also pointing to a cost and it's not a hypothetical one. The NEU's 2026 findings on declining critical thinking are corroborated by what teachers are reporting directly - pupils using AI to "do their homework" rather than to think through a problem, a growing reliance on AI for tasks that used to require pupils to draft, revise and struggle a little before arriving at an answer. One secondary teacher's comment, reported in coverage of the survey, captures it plainly, "Staff are not trained to use it properly, but are using it and it's producing sub-standard slop."

There's also a more structural concern and it's the one we think deserves more attention than it currently gets. Only 47% of pupils, according to research cited in the State of the Nation - AI in Education report published in May 2026, feel confident judging whether AI-generated content is actually accurate. That's not a media literacy footnote, that's nearly half of pupils unable to reliably tell good information from confidently-stated nonsense at exactly the point in their education when learning to make that judgement matters most. And the access picture isn't equal either as the same report found that only 21% of state school teachers have received formal AI training, compared with 45% in private schools. If AI in education becomes a story about which schools can afford to deploy it thoughtfully and which can't, that's a problem that compounds existing inequality rather than solving it.

Nearly half of schools (49%) currently have no AI policy whatsoever, for staff or pupils. Two-thirds have no policy specific to pupil use, and that's not a sector that has decided AI is safe and well-managed, it's a sector where adoption has significantly outpaced governance.

The thing both "sides" of this conversation tend to skip

The debate around AI in schools often gets framed as a binary, either you think AI is a long-overdue solution to an unsustainable workload crisis, or you think it's eroding the thinking and human connection that education is supposed to build. Both framings, on their own, miss something important.

The workload-relief framing tends to treat "time saved" as the only metric that matters, without asking what's being given up in exchange when a task that used to require a teacher's professional judgement: deciding what a particular pupil needs this week, crafting feedback that responds to how that specific child is thinking - gets handed to a tool that has no actual knowledge of that pupil.

The erosion framing, meanwhile, sometimes underestimates how real and severe the underlying workload problem is, and risks romanticising a status quo that was, by every available measure, already failing a significant proportion of the profession before AI entered the picture.

The honest position, we think, sits between those, AI can be a legitimate and valuable tool when it's used to remove genuinely low-judgement burden from teachers' weeks and it becomes a problem the moment it starts substituting for the judgement, relationship-building and thinking that are the actual substance of teaching and learning, not the administrative scaffolding around it.

The Department for Education's own framing, in its recent guidance, gestures at this distinction - AI tools should support a "manageable workload," but the policy explicitly reaffirms that "teachers remain at the heart of classroom practice" and that it is "education professionals who determine how and when technology can enhance learning." That's the right instinct, but the harder part is turning it into something more concrete than a guiding principle - actual policies, actual training and actual decisions about where the line sits in a specific school, for specific tasks.

Questions worth asking before adopting any AI tool in your school

If your school is weighing up where AI genuinely belongs in your workflow, these are worth working through directly, ideally with staff involved in the conversation rather than having a tool introduced to them after the decision is made:

• Does this remove low-judgement burden, or does it remove judgement that should remain with a teacher? Drafting a first version of a newsletter is different from a tool deciding what feedback a struggling pupil receives on their writing;
• Will pupils understand that AI-generated content can be wrong and how to check it? If the answer is no, that's a literacy gap to address before the tool becomes routine, not after;
• Is there a human reviewing the output before it reaches a pupil or parent? Under the DUAA's requirements around automated decision-making, this isn't just good practice, for any process making a decision that significantly affects an individual, a path to human review is now a legal expectation;
• Does your school have a policy at all? With nearly half of schools currently without one, this is the single most overdue conversation for many leadership teams to have, regardless of which tools you eventually decide to use;
• Who has access to training, and is that access equal across your staff? A tool is only as good as the confidence and skill of the person using it, and right now, that confidence is unevenly distributed across the sector.

Where we land on this

We created technology for schools, so it would be easy for us to simply champion every tool that promises to make a teacher's week easier. We don't think that would be honest and we don't think it would actually serve the schools we work with.

AI has a real, evidenced role to play in reducing the administrative weight that's been crushing the profession for years and pretending otherwise does teachers no favours. But the parts of teaching that matter most: judgement, relationship, the slow work of helping a child think something through rather than be handed an answer - aren't burdens to be automated away.

The schools getting this right aren't the ones moving fastest and they're not the ones refusing to engage with AI at all, they're the ones being deliberate, clear about which tasks genuinely benefit from AI's speed, and equally clear about which tasks need a human mind, paying full attention, doing the slow work that technology can't shortcut.