Ken Jones is Emeritus Professor at Goldsmiths, University of London. He writes about the politics of education, in Britain and in Europe. He has been an English teacher, a union activist and, until 2024, a member of the education policy team at the NEU.
Google CEO Sundar Pichai thinks that AI is ‘probably the most important thing that humanity has ever worked on, more profound than electricity or fire’. Yann LeCun, for ten years chief AI scientist at Meta, compares it to the invention of the printing press. Bridget Phillipson agrees with him: AI is the ‘most radical force for progress since the 15th century’. For the Tony Blair Institute, recipient of more than £200 million from the founder of the Big Tech company Oracle, AI is a transformative and empowering force that allows us to ‘reimagine the state’, education included.
Educators are more grounded in their assessment. Rather than being enthralled by imagined futures, they are concerned with AI technologies as they actually exist now in educational settings, under-funded, fragmented and over-managed. The NEU’s evidence on AI to the House of Commons Education Committee speaks of ‘funding shortfalls’, ‘policy churn’ and ‘unregulated and piecemeal change’; in this context, the Union argues, it is difficult for educators to respond to AI technologies with a clear sense of educational purpose.
Researchers agree: AI has the potential to ‘extend and augment how teachers want to teach their students’. But realising this potential requires conditions that don’t yet exist: technology designed to centre educational values; autonomy for teachers to decide how best to use it; and time for professional development.
Innovation
AI is often represented as a disruptor of systems, an innovative force. The NEU’s ‘State of Education Survey 2026’ shows that there is some truth in this. Some of the working and learning practices of educators and students are changing rapidly. The number of teachers using AI for resource creation has risen beyond 50%. Nearly half of primary and Special/PRU teachers have used AI for lesson planning, and a third of secondary teachers. Survey respondents also noted that students’ working habits are being radically changed by AI, through its extensive use for homework ‘whenever they can really’.
But this is not the whole story. In many ways, AI’s arrival in schools and colleges is not a new dawn for education but a moment of repetition, in which already-existing trends are reinforced. AI technologies are not free-standing; they are attached to established practices that involve the prescription of curricula, lesson planning and assessment schemes.
An absorbing recent study shows this process of attachment and reinforcement at work. Louise Couceiro and her colleagues get under the skin of a widely used maths platform, demonstrating how it weaves the fabric of classroom life into a pattern where pedagogic practices, student-teacher relationships and students’ perceptions of themselves and their peers are inter-related, and the central motif is not ‘quality of understanding’ but task completion – getting it done’. The platform is ‘less a pedagogical aid’ than a ‘mechanism of behavioural oversight’.
Bridget Phillipson says that the DfE ‘sees the opportunities of AI to superpower the learning of every child – especially children from disadvantaged backgrounds. Her department has recently asked edtech and AI companies to submit tenders for an ‘AI Tutoring Tools Pioneer Programme’. What the DfE is asking for looks very like the technology that Couceiro and her colleagues are criticising: ‘a chatbot that can answer curriculum-aligned questions’; ‘a question-generation engine that produces curriculum-aligned practice questions, and a ‘system for tracking and synthesising student progress.’ This is what policy-makers call ‘personalised learning’, but in reality, it turns children into ‘thin subjects’, who have ‘become their performance’, their identities diminished into data points.
Education is social
Teachers who responded to the NEU survey are very clear on the fundamental problems of ‘personalised learning’ as it is being realised in the classrooms described by Couceiro, and in the isolation rooms of internal exclusion systems. Only 14% of them agreed with the DfE’s plans, with teachers working in Special Education/PRUs being the most strongly opposed. The terms of their criticism are significant: it is not the technical functioning of personalised learning tools that they single out; it is rather their impact on teaching skills and human interaction. They see teaching as a profoundly social activity and AI as a potential agent of its desocialisation.
In this, their thinking is informed by the experience of the pandemic, from which they draw very different lessons from the government. During Covid-19, IT-mediated learning met an immediate need, but it was at best what one teacher in the survey called ‘a blunt instrument’. Yet in the years since 2021, governments have poured resources into online learning, especially in the form of Oak, without doing anything like enough to attend to other dimensions of the social crisis that the pandemic had highlighted – notably those of mental health and student engagement. These problems cannot be dealt with by ‘personalised learning’ as the DfE imagines it. The Curriculum and Assessment Review may speak the language of engagement, but the practices towards which schools are guided by the demands of accountability point in a different direction.
Teachers’ work
In her book, Teaching Machines, Audrey Watters reminds us of the promises made by those who introduced mechanical teaching devices into the classrooms of the 1920s: they spoke in ways almost identical to those used by the Tony Blair Institute today, summoning up a vision of the school in which a teacher could focus on her ‘real function, inspirational and thought-stimulating activities’. From then until now, a widening stream of thought, passing through the behaviourist psychology of Fred Keller and B.F. Skinner, has promoted a model of learning and teaching in which step-by-step mastery of bite-sized content, supported by individualised instruction, whether human or mechanical, is seen as the way forward for the school.
We don’t need a crystal ball to see how such visions will be realised. We just need to look hard at where we are now. Teachers are already working with online data supplied by personalised learning systems and records of attendance. Many of them are using curriculum planning resources supplied by Oak, through which they are guided by Oak’s AI lesson assistant, Aila. For in-school assessment, any number of tools are available, starting with Reception Baseline Assessment. For behaviour management and home-school communications, teachers use Dojo or a tool very like it.
In other words, generating and managing data have become integral to teachers’ work. Their activity is increasingly steered by AI tools, which also provide the data that compares the effectiveness of their work to that of other education workers. To borrow a phrase from the French economist Cédric Durand, AI is becoming the ‘general intellect’ of schooling, the means by which the work of teachers is co-ordinated.
This raises fundamental questions. What expertise and authority can teachers lay claim to, when core aspects of their work – determining curriculum content, making pedagogic choices – are passed over to algorithmic tools, whose ‘decisions’ are not transparent? What is left of teacher professionalism when principles of autonomy and collective agency are diminished? The Tony Blair Institute promises teachers a future as ‘mentors and learning designers, supported by AI co-pilots to give meaningful feedback and adapt teaching to need’. The teachers quoted in the NEU survey, like many classroom researchers, are sceptical.
Organising
How can the brake be applied to these accelerating tendencies? How, if at all, can AI in education be turned in a different direction?
Trying to hold in check the process of ‘unregulated and piecemeal change’ that the NEU has criticised, the DfE is backing the EdTech Evidence Board, which aims to support schools in their ‘critical thinking’ about AI procurement. In a situation where schools and trusts are making their own, sometimes uninformed, decisions about what AI tools to adopt and how to use them, this is a useful if limited response.
The NEU has called for something more – a national EdTech strategy, developed in partnership with education unions’. This is an important demand, because it potentially broadens discussion of AI, beyond issues of procurement to questions of the educational purposes which AI is meant to serve.
What the Union is demanding at a national level needs also to be pressed for in schools, where educators are experiencing the effect of decisions about AI use which they were not involved in making. Working conditions are being transformed, but the tools of transformation are encased in ‘black boxes’ of algorithmic decision-making which are not subject to enquiry and challenge. Likewise, what learners do with learning platforms, and what learning platforms do to them, is seldom investigated, despite the disquiet of educators documented by the 2026 Survey.
The time is right for educators to carry out what Couceiro calls ‘technological audits’ of their workplaces to evaluate the effects of AI, against criteria that are explicitly educational, and to work with resources that embody principles which they can confidently support. It is no exaggeration to say that developing knowledge and expertise around these questions is now essential to defending teachers’ claims to professional status and autonomy. Here the Union has a major new role to play.
References
Bridget Phillipson’s ideas about AI in education can be found here. The Tony Blair Institute has set out its programme in several statements.
The article by Louise Couceiro, Rebecca Eynon and Laura Hakimi is published open access in Anthropology & Education. Rebecca Eynon’s evidence to the House of Commons Education Committee on AI is online, alongside the Union’s own evidence.
Information is available about the government-supported Ed Tech Board and the DfE’s invitation to edtech companies to bid for funding to build personalised learning tools. Also read The Education Policy Institute’s report on MAT leaders and AI implementation.
Audrey Watters’ book Teaching Machines is published by MIT Press. Cédric Durand’s article ‘After AI’ appeared on the Sidecar site. The idea of children as ‘thin subjects’ is borrowed from an article by Stephen Ball and Emiliano Grimaldi.
The NEU commissioned 2025 research on standardised curricula.
Finally, a mention for the journal Learning, Media and Technology, an essential reference point.