The Case for Subject-Specific AI Training

Last term, I ran two CPD sessions on consecutive days at the same school. On Monday, I worked with the English department. On Tuesday, the science department. The content was about AI integration in teaching; the conversations were completely different.

The English teachers wanted to talk about academic integrity. How do you set an essay when students have access to ChatGPT? How do you assess creative writing when AI can generate passable prose? How do you teach critical analysis when students can ask an AI to critique a text for them? Their concerns were existential, AI does not just affect how they teach, it challenges what they teach.

The science teachers wanted to talk about efficiency and accuracy. Can AI help generate lab reports at different ability levels? Can it create model answers for exam-style questions? Can it produce differentiated data sets for analysis tasks? Their concerns were practical, AI is a tool that can save time and improve resource quality, and they wanted to know how to use it well.

Both groups are right. Both sets of concerns are legitimate. And both groups would have been badly served by a generic, whole-school AI training session that treated all subjects the same.

The Problem with One-Size-Fits-All AI CPD

Most school AI CPD is delivered whole-school. Everyone sits in the same room, hears the same presentation, and works through the same activities. The rationale is efficiency: one session, all staff, job done.

But AI does not affect all subjects equally. It does not raise the same questions, create the same risks, or offer the same opportunities across disciplines. Treating it as though it does wastes time for everyone: the English teacher who needs to rethink assessment from first principles sits through a demonstration of AI-generated worksheets they will never use, while the science teacher who wants practical tool guidance sits through a philosophical discussion about authorship that feels irrelevant to their practice.

The result is disengagement. Teachers leave feeling that the CPD was "not really about my subject," and they return to their classrooms without any meaningful change to their practice.

How AI Impacts Subjects Differently

Let me map out the differences across a few subject areas, because the variation is more significant than most school leaders realise.

English and humanities. These subjects are at the sharp end of the AI challenge. The core skills they teach, writing, argumentation, critical analysis, creative expression, are precisely the skills that AI can simulate most convincingly. An AI-generated essay is far harder to distinguish from a student-written essay than an AI-generated lab report is from a student-written one. This makes assessment integrity a central concern.

But the opportunity is equally significant. AI can be a powerful teaching tool in English when used as a sparring partner rather than a ghostwriter. Students can generate AI text and then critique it. They can compare AI analysis of a poem with their own. They can use AI to explore alternative perspectives on a text. The pedagogical possibilities are rich, but they require a fundamentally different approach from the way English has traditionally been taught and assessed.

Sciences. AI is primarily a productivity and resource tool in science. It excels at generating differentiated worksheets, creating model answers, producing data sets for analysis tasks, and explaining complex concepts at different levels of accessibility. The assessment integrity risk is lower because science assessment relies heavily on practical work, data analysis, and structured problem-solving that are harder for AI to replicate convincingly.

The specific AI considerations for science are more about accuracy and reliability. AI language models can and do produce scientifically inaccurate content; they may describe chemical reactions incorrectly, misstate physical laws, or generate plausible-sounding but wrong explanations. Science teachers need to be particularly skilled at verification, and their students need to understand that AI output is not equivalent to a peer-reviewed source.

Mathematics. AI's relationship with mathematics is evolving rapidly. Current language models are inconsistent with mathematical reasoning; they can solve some problems correctly while making basic errors on others. But specialised AI tools for mathematics are improving quickly, and the question for maths teachers is shifting from "can AI do this?" to "what does it mean for our teaching when it can?"

The CPD needs for maths teachers include understanding the capabilities and limitations of AI mathematical reasoning, exploring how AI tools can support differentiation and personalised practice, and rethinking assessment in a world where step-by-step problem solutions can be AI-generated.

Languages. AI is arguably most transformative in language teaching. Translation tools, conversation practice bots, grammar checkers, and pronunciation feedback tools are all areas where AI can genuinely support language acquisition. The CPD needs here are about integration: how to use AI tools alongside traditional teaching methods to accelerate learning; and about discernment, helping students understand that AI translation is a scaffold, not a substitute for genuine language competency.

Creative arts. AI image generation, music composition, and creative writing tools raise profound questions about originality, authorship, and the purpose of creative education. Arts teachers need CPD that engages with these philosophical questions alongside practical tool training. The conversations I have had with art and music teachers are often the most thoughtful and nuanced of any department.

What Subject-Specific AI CPD Looks Like

Effective subject-specific AI CPD has three components.

Discipline-specific use cases. Show teachers how AI applies to their specific subject. English teachers need to see AI used for textual analysis and writing instruction. Science teachers need to see AI used for resource generation and data creation. Maths teachers need to see AI used for differentiated problem sets and worked examples. Generic demonstrations do not land.

Discipline-specific risks. Address the challenges that are unique to each subject. English needs deep engagement with assessment integrity. Science needs focus on factual accuracy and verification. Arts needs engagement with questions of originality and authorship. Whole-school training cannot go deep enough on any of these.

Discipline-specific assessment redesign. Every department needs to rethink its assessment approach in light of AI, but the redesign looks different in each subject. English might move towards more in-class writing, oral assessment, and process-based portfolios. Science might increase the weight of practical assessment and data analysis. Maths might incorporate more oral explanation of reasoning. These are subject-level decisions that require subject-level CPD.

How to Structure It

I recommend a model that combines whole-school and department-level CPD.

Start whole-school for the foundations: what AI is, how it works, the school's AI policy, data protection principles, and the shared ethical framework. This gives everyone a common language and a shared baseline.

Then move to departments for application, risk assessment, and assessment redesign. Give each department a facilitated session, or better, a series of sessions, focused on their specific subject context. If you have an AI lead or a CPD coordinator, they should work with heads of department to design these sessions collaboratively.

Reconnect whole-school for sharing. Once departments have done their subject-specific work, create opportunities for cross-departmental sharing. An English teacher's approach to assessment integrity might inform the history department's thinking. A science teacher's approach to AI-generated resources might be adapted for geography. The cross-pollination is valuable, but it works best when it builds on subject-specific depth rather than replacing it.

The Investment Case

Subject-specific CPD is more expensive than whole-school training. It requires more sessions, more facilitation time, and more planning. School leaders rightly ask whether the additional investment is justified.

I believe it is, and here is why. Whole-school AI CPD that does not change classroom practice is not cheap, it is wasteful. The time, the speaker fees, the supply cover, these are real costs. If teachers leave without applying what they learned because the content was not relevant to their subject, that investment is lost.

Subject-specific CPD costs more upfront but produces measurably better outcomes. Teachers apply what they learn because it connects directly to their daily practice. Assessment redesign actually happens because it is grounded in subject-level reality. And the school's overall AI integration is more coherent because each department is working through the implications for their discipline rather than applying a one-size-fits-all template.

The AI Literacy Audit Tool assesses teacher competency across the school, and in my experience, schools that invest in subject-specific CPD show faster and more consistent improvement across the audit's teacher competency dimension. The data supports what common sense suggests: teachers learn best when CPD speaks to their professional reality.

If your CPD budget can stretch to one thing this year, make it subject-specific. Your teachers deserve better than a generic session that leaves them wondering how any of it applies to their classroom.