What Do We Do Next

In my previous post, I argued that ‘curriculum’ is a floating signifier whose meaning has been temporarily fixed through the ‘knowledge turn’ in English education and that LLMs are circumventing the traditional mechanisms through which meaning is selected, contested, and stabilised: the conference paper, the inspection framework, the trust CPD programme, the slow capture of professional common sense. The question I was then asked, by more than one person I respect, was the obvious one. Fine. But what do we do next?

What follows is an attempt to answer this question honestly; I do not have a complete answer. If someone, or an organisation, tells you otherwise, I would ask what they are selling.

There are, essentially, two ways to respond. The first is direct and frontal: ban AI, restrict access, install detection software, enforce compliance. It is fast, visible, and satisfying. It is also, in the current context, as futile as commanding the tide to recede. Students have AI on their phones. Teachers are using it to plan lessons. The encoding is already hybridised. The second is slower and harder and involves the sustained work of shifting assumptions, language, and institutional common sense within which decisions are made. The ‘knowledge turn’ did exactly this by fixing what curriculum meant in the professional imagination, so that ‘knowledge-rich’ and ‘powerful knowledge’ became common sense and alternatives became unthinkable.

Antonio Gramsci called these two kinds of responses a war of manoeuvre and a war of position. Schools currently fighting a war of manoeuvre cannot win because they do not have the resources or time to challenge global corporations and the ecosystem around them directly. Moreover, they would be fighting on the terrain that has been set by what they want to challenge. My view is that the appropriate response is a Gramscian war of position: a change in how the profession thinks about the curriculum, what a teacher does, and what it means for knowledge to be authoritative. For it to work, it needs to be built into the culture of schools and into the training of new teachers.

It would be very reasonable to point out that this would take years while model updates take seconds and are supported by advertising. The Gramscian war of position does not seek to match the pace of technological change and claims something more modest: that the alternative is a faster way to lose. A profession that responds only through a direct challenge does not slow the new settlement; it simply ratifies it by failing to contest the frame. My suggestion is one where the profession retains agency over what curriculum means, rather than handing that question to bodies with no accountability to students, teachers, or the public.

Adopting this frame requires a counter-hegemonic strategy. What this means is a challenge to the dominant framing of LLMs in education as neutral tools for efficiency, personalisation, and workforce preparation. That framing is not a neutral observation. It is the position of technology companies with significant commercial interests in education, international bodies whose policy frameworks have absorbed the efficiency argument, and a DfE whose guidance has moved in that direction. This framing can be seen in trust policies, edtech marketing, OECD reports, and the professional development programmes that schools are rolling out. It offers a new common sense that is algorithmically encoded, smoothed and unattributed, making it harder to contest than the one it seeks to displace.

The first move, then, is to make that common sense visible as a position rather than as a fact. Each of the following suggestions is designed to make the assumptions about curriculum, authorship, and authority visible, and therefore contestable.

Map the hybrid encoding. Ask every department to identify where LLMs are already being used to produce curriculum materials. Most schools have no idea how much of their enacted curriculum is co-authored by LLMs. If significant proportions of the materials in use are generated by LLMs, then the curriculum-as-enacted is not the curriculum-as-designed. I would caution here that a focus on individual disclosure creates an incentive for concealment so anonymous or aggregated departmental mapping, focused on materials rather than individuals, would help produce more information in what should be clearly named an audit, not a disciplinary process.

Distinguish between AI-assisted and AI-generated.

AI-assisted means the teacher reviewed, revised, and took professional ownership of the output, with the LLM encoding passing through teacher judgement. AI-generated means the output went from the model to the student without that mediation. Most schools make no distinction between these two uses and treat them as the same, missing that teachers are most likely to use LLM-generated materials without review if they are under pressure and still-developing subject and pedagogic knowledge. Implementing this distinction requires pairing it with support rather than surveillance, with time for review, collaborative planning, and CPD with practical outcomes included.

Test the displacement.

Choose a topic a department is about to teach and put the questions students would ask to two or three LLMs. Compare the responses with the planned sequence. Where do they differ? What does the LLM include that the teacher deliberately defers? Where does the LLM present as settled what the curriculum treats as contested? This exercise is only as good as the subject knowledge brought to it, which is itself an argument for doing it as a departmental conversation where it strengthens curriculum planning and develops subject knowledge in the process.

Show the LLM encoding explicitly.

Open the lesson by showing the LLM response to the topic/question on the board. Then ask: what has it included? What has it left out? Where does it differ from what we have been building? Why might those differences matter? This makes the epistemic work visible rather than leaving it implicit.

Ask the right governance questions.

Most governing bodies, when they engage with LLMs at all, will normally ask whether the school has a policy and whether it has been followed. These are the wrong questions. More productive questions would be framed as follows: What proportion of the materials students encountered this term were produced by the institution, and what proportion by LLM systems it does not control? What review process was applied to LLM-generated materials? Does the assessment framework still correspond to what students have actually encountered? These questions require governors to understand that the curriculum-as-assessed may no longer match the curriculum-as-encountered, and that this mismatch is a strategic risk for students, the institution and teaching staff.

One question that was raised was what should happen in the education of teachers. The current ITE curriculum teaches trainees to plan, sequence, and assess. It teaches them to be encoders and, to some extent, stabilisers, though it does not use that language. If epistemic orientation is to become a professional competency, it has to be built into teacher formation as a foundational element, not a module on digital literacy. ITE programmes need to do three things. First, establish that the teacher is a stabiliser of meaning, not merely a deliverer of content, and that if this function is outsourced it is simply performed elsewhere by systems the institution does not control. Second, be explicit that LLMs are a co-encoder of the curriculum students encounter, operating outside the institution’s control and often its awareness. Third, build the capacity to review LLM-generated materials not only for accuracy but for encoding: what has been foregrounded, smoothed, or left out, and whether the result serves the epistemic commitments of the curriculum or quietly works against them.

None of the above is a series of quick wins or a policy template. They are the slow work of changing what the profession takes for granted and it is worth remembering that the ‘knowledge turn’ took the better part of a decade to move from the arguments of the few to the common sense of the many. It won because it captured the institutional mechanisms of inspection, CPD, trust culture, and teacher training. Epistemic orientation will need to do the same, and it requires institutional allies willing to examine their current position rather than simply manage it. I would rather a profession, leadership and governance that makes decisions consciously, not because it is easier, but because that is what professional responsibility in a democratic institution as part of civil society requires. The alternative is to simply continue to refine the institutional encoding while the site of curriculum meaning shifts quietly elsewhere until it becomes a crisis that requires urgent manoeuvre rather than a sustained position.