The Return of the Specialist: What AI Means for Teaching Expertise

In almost every school I visit, there is a teacher whose classroom is the epitome of good practice: great delivery, students engaged, genuinely curious, asking questions that catch you off guard. Yet that same teacher might nevertheless be horrified when asked to update a spreadsheet or navigate the MIS in search of data. For years, we have treated this as a gap to be filled. We have enrolled them on courses, written targets around their digital competence, and measured their progress against frameworks that positioned technological fluency as a baseline expectation.

I have been wondering lately whether we were solving the wrong problem.

The TPACK model, developed in the early 2000s, was an attempt to make this expectation intellectually coherent. It proposed that effective teaching required the meaningful overlap of three knowledge domains: content, pedagogy, and technology. At the intersection of all three sat the ideal practitioner. It was a reasonable framework and, in its way, an honest diagnosis of what schools were being asked to produce.

The difficulty is that most people’s circles aren’t equal in size, and often they don’t overlap neatly in the centre.

In practice, communities compensate. The tech-confident colleague covers for the subject expert who cannot work the HDMI input switch. The outstanding classroom practitioner leans on the data manager to make sense of the spreadsheet. Schools function because individuals with different strengths find ways to work around each other’s gaps. This is how most professional communities actually operate, and it works reasonably well when leadership is strong enough to make the complementarity intentional rather than accidental.

But it does raise a question that I think deserves more attention: was the expectation of full-circle overlap ever realistic, and more importantly, is it necessary?

Recently I spent an afternoon building a chatbot for my website: a version of me, trained on my writing and thinking, that visitors can question without my being present. Have a play, it on the bottom right corner of your screen. Before that, I had put together an audit tool that allows school leaders to work through a structured set of questions and receive a personalised framework for their digital strategy.

In both cases, the domain expertise (the knowledge of what good digital strategy looks like, the understanding of how schools function, and the years of accumulated experience that sit behind the questions I ask) was entirely mine. What AI provided was the mechanism. The construction. The part that, not that long ago, would have required a developer and considerable expense.

I found myself thinking back to the mid-2010s, when I was responsible for digital strategy across a group of three schools, overseeing the rollout of a couple of thousand one-to-one devices to students and staff. That work put me at the intersection of technology, pedagogy, organisational change, and leadership in ways that were genuinely demanding. I was stretching into circles that were not naturally mine.

This group of schools had built a bespoke management information system that was beloved by their staff but required a team of specialist coders to make it work and to keep it running and up to date. The knowledge required to specify what the system needed to do had always existed in the school. However, the knowledge required to build it had to be bought in and maintained, again, at considerable expense.

That gap is closing fast, and I find myself asking what happens when it has closed altogether.

The standard narrative about AI and professional development runs roughly like this: as technology becomes more powerful, professionals need to broaden their competencies to keep pace. The circles need to get bigger and to overlap more. These days it would seem that, on top of everything else, everyone needs to become, to some degree, a prompt engineer, a data analyst, and a digitally fluent operator who can navigate an increasingly confusing AI landscape.

I wonder whether this is exactly backwards.

If AI can absorb the technical layer, as it did when I built the audit tool and the chatbot, and if domain expertise can be translated into functional tools without requiring the expert to also be a developer, then the logic inverts. What matters is no longer the ability to sit at the intersection of multiple circles, but rather the depth of the circle you most naturally inhabit.

The teacher who understands learning, who knows their subject, who has spent years reading students in a classroom, becomes more valuable in that reading, not less. The leader who can read an organisation, ask the right questions, and make sound judgements about what a school needs does not also need to know how to configure a database.

A fair challenge to this argument would be that I am reasoning from my own experience, and I am aware that I arrived at those afternoon projects with years of accumulated context behind me. The chatbot and the audit tool work not because AI is straightforwardly democratising, but because I was able to bring enough domain depth to the process to direct it meaningfully. Someone without that depth would likely produce something that looked functional but wasn’t. So the argument then is not that AI removes the need for expertise. It is that AI may be challenging where the expertise needs to sit. Away from the technical function, and back towards the domain.

In this light, the TPACK model was perhaps not a vision of what professional expertise should permanently look like, but rather a description of a transitional moment: a period when engaging with technology meaningfully required a kind of fluency that did not always come naturally to those with the skills more usually associated with great teaching or leading.

But now that technology is useful, uncomplicated, and accessible to almost anyone, is that moment ending?

If it is ending, then the implications may be significant. A school system that spent two decades telling teachers they needed to become more digitally fluent might need to revisit what it was actually asking them to become fluent in, and why. If the technology layer is becoming something that responds to expertise rather than demanding it, then the professional development conversation changes. The question shifts from “how do we get everyone to the intersection?” to “how do we deepen the circles that matter most?”

There is something in this for curriculum too, though I hold that thought loosely and I am wary of overreaching.

There is a version of this story that is straightforwardly encouraging: an AI-rich terrain frees professionals to specialise, to deepen, to be genuinely excellent at the things they are actually good at. The extraordinary teacher who struggles with the MIS gets to be extraordinary without apology. The leader with sound strategic instincts does not also need to understand the technical architecture behind the systems they use.

Whether that version of events is what actually unfolds will depend on choices that schools, systems, and policymakers have not yet fully confronted. Look at me with my chatbots. The technology seems to be running ahead of the thinking, as it usually does.

What I am more confident about is this: the assumption that wider circles and more overlap is always better deserves scrutiny. The communities that work best are not necessarily the ones where everyone spans everything. They are the ones where different kinds of depth are genuinely valued, and where the organisation is structured to make the most of them.

I would be interested to know whether that resonates with what others are seeing.

Photo by kaffie.co on Unsplash