Sylvain Savier is a secondary school teacher and an NEU rep and officer in Redbridge District.
When I was asked to present at the NEU AI workshop in February, I started by showing colleagues three bullet points from a real school AI policy, and I asked: what's good here, and what's missing?
- Staff must remain accountable for AI-generated content;
- Specific permission is required for sensitive data;
- Use AI transparently and ethically.
The answers came quickly. What's missing is power. Who wrote this policy? Who was consulted? Who benefits? A policy that tells teachers they are "accountable" for AI outputs while giving them no meaningful say in which tools are adopted, or how, is a transfer of liability, not a meaningful safeguard. Another answer highlighted how magical thinking disguises itself as pragmatism: "Delivering efficiencies", "improving effectiveness", "empowering decision making" — performative language.
In a different society, we would be asking how AI best fits and reinforces our education system. Instead, school communities do what they know best in the education marketplace: they "embrace change", tinker with opportunities, and hope for the best while AI quietly rewires how pupils actually think.
Our objective is to move from policy to power. But it is a titanic struggle when policy is being designed on the ever-shifting ground of Silicon Valley, and schools are entering commercial agreements about products that will be unrecognisable within eighteen months.
The problem with magical thinking
Magical thinking operates on two levels. The first is familiar. Some call it proletarianisation — when you end up working to the tune of technology, as opposed to the machine freeing you for more creative or socially rewarding work. Almost every school AI policy begins with the same alleged opportunities: efficiencies, reduced workload, lesson planning done in seconds. Compelling. But efficiency — and indeed, productivity gains — only materialise if the time saved is actually returned to teachers. Too often it isn't. The NEU's own guidance warns explicitly that teachers' work risks being "displaced and shifted towards more routine tasks and the administration of the technology itself".
The second level concerns the logical outcome of optimisation by AI — something better known as the dialectic. Optimisation — and all the nuts and bolts of EdTech consultants (standards, performance, cost reduction) — sounds common sense. Who wants inefficiency after all? But optimisation, taken to its AI logical endpoint, is not efficiency. It is the deletion of jobs. What happens to the people whose roles are "optimised away" does not get answered by a policy document. It gets answered — or not — by the society we build. As a union, we must lead the fundamental rethinking of the future of work in people's lives that AI is forcing upon us. This must include pushing back against language that perpetuates magical thinking – the language of EdTech – within NEU-negotiated AI policies.
Something big Is already happening
None of the above is theoretical. In the same week as the NEU workshop, Matt Shumer, founder of an AI company, published a piece aimed at people outside the tech industry. His message: the disruption has already happened to people like him, and everyone else is next. He described handing a complex project to an AI, returning hours later to find it completed better than he could have in days. Meanwhile, Anthropic's CEO has predicted AI will eliminate 50% of entry-level white-collar jobs within one to five years. Shumer calls this estimate conservative.
As French philosopher and economist Frédéric Lordon writes, this is “the dialectic of AI”. AI is now systematically dismantling the "creative bourgeoisie" that prospered in the “knowledge economy”. Capital is eating the very class that, unconcerned, endorsed previous waves of automation while workers in other sectors were thrown onto the scrapheap by offshoring, robotisation and zero-hours contracts. A similar indifference is visible at school: we are equipping students for a job market without seriously asking what that market will look like, or whether "employability" should remain (ever have been) education's organising purpose.
This is the context in which policies are being written. But the window of opportunity for value extraction is fast closing for the EdTech companies themselves. Chegg, once a dominant homework platform, lost 99% of its market value and cut nearly half its workforce after ChatGPT replaced its core product overnight. If we are not to simply co-manage an unplanned end of work, the euphemisms — "efficiencies," "raised standards," "improved outcomes" — must be recognised in policy making as the acceptable face of a much more aggressive underlying logic.
What good policy can and cannot do
Some schools are genuinely trying. Woodbridge High School's policy is better than most: it names approved tools, sets age restrictions, handles data carefully, exemplifies acceptable use cases and states that AI "must not make autonomous decisions regarding student discipline, marking or support without teacher oversight." Union involvement made a visible difference in framing AI as a means to "reduce workload" — not to extract more output from existing staff. That distinction does not appear by accident. It appears when union members are in the room.
But even the best policy operates within the logic of optimisation. No policy can address what happens when AI tools become so capable that the question is no longer “how do we use AI well?” but “how many of these roles do we still need?”
Technology won’t wait; neither should we
Historically, we have defended the conditions of work. We may now need to defend its existence — or argue seriously for what replaces work when AI makes it scarcer.
What does the optimisation of education actually mean? Who benefits? Who bears the cost, and what kind of society do we want when this wave of automation has run its course?
Good school policy is a start, but staff must be at the table, or they remain on the menu. If we’re not immediately going to get the left-wing government needed to start these conversations systematically, then start them with your members, your headteacher, your Local Authority now. The table won’t come to you.