Schools need more than AI tools. They need learning systems.

Adam Sturdee
Jun 25
7 min read

What Satya Nadella’s thinking on AI means for the future of schools

Satya Nadella has made an argument that should matter to every school leader thinking seriously about artificial intelligence.

His point was not that AI is powerful. That much is obvious. Nor was it that every organisation should start using it, since most already are, formally or otherwise. His deeper claim is that the future will belong to organisations that build their own learning systems around AI, rather than handing their knowledge, judgement and processes to a small number of frontier models.

That distinction matters for schools.

In a long note , titled “A frontier without an ecosystem is not stable,” Nadella warns against a world in which a handful of dominant models absorb the expertise of everyone else. If that happens, he argues, entire sectors risk being hollowed out in the way some industrial economies were hollowed out by globalisation. Organisations become users of intelligence rather than builders of it. They feed the system without ever owning the loop.

For schools, this is a leadership question before it is a technology one. Are we using AI as a collection of tools, or are we beginning to build institutional intelligence?

Using AI and building with AI are not the same thing

Most of the conversation about AI in schools is still tool-led. Which chatbot should staff use. Which marking tool is safest. Can we automate reports, generate resources, stop pupils cheating. These are reasonable questions. They are practical, immediate and often urgent. They are also not strategic enough.

A school that adopts AI tools may become slightly more efficient. A school that builds intelligent systems around its own professional knowledge can become genuinely more adaptive. That is the shift Nadella is pointing towards.

In business terms he describes a relationship between human capital and token capital. Human capital is the knowledge, judgement, creativity, relationships and pattern recognition that people bring. Token capital is the AI capability an organisation builds and owns once that knowledge is captured, structured, reused and improved over time. The two are not in competition. As he puts it, human capital becomes more valuable as token capital grows, because without human direction you have compute running in circles.

In schools, human capital is everywhere. It is the experienced teacher who knows exactly when to pause before taking an answer. It shows in the head of year who can read the temperature of a corridor before anything has happened, and in the subject leader who can name, in seconds, why a misconception has taken root. The question is whether AI helps schools preserve and scale that judgement, or quietly flattens it.

I have written before about how school leaders themselves should already be using AI to think, decide and lead. The same logic now applies at the level of the whole institution.

The risk is that schools become consumers of generic intelligence

There is a real risk that schools become passive consumers of generic AI. A teacher asks for a lesson plan, a leader for a policy, a pupil for an explanation, a trust for a strategy document. The output can be fluent, fast and impressive, and still strangely thin, because it is not rooted in the actual context of the school.

Schools are not generic organisations. They are living cultures, with histories, routines, tensions, inherited language and deeply local knowledge. A behaviour policy that works in one school can fail in another. A coaching model that feels empowering in one culture can feel threatening in the next. AI that does not understand this can produce the appearance of wisdom without the substance of it.

We have been here before, with technology that promised transformation and delivered drawers full of unused logins. The lesson is that the tool is rarely the point. What matters is whether it strengthens what the school already does well.

The opportunity is schools that learn from themselves

The better future is more interesting. Picture a school where AI does not replace professional judgement but makes it more visible. Staff development informed by real patterns in classroom practice rather than occasional snapshots. Policies that improve because leaders can see how they actually play out in the lived experience of staff and pupils. Professional learning built around the genuine questions emerging from classrooms, rather than guesswork.

This is where Nadella’s thinking becomes useful for education. The school of the future is not one where machines think for teachers. It is one where human expertise and AI capability form a responsible learning loop. Teachers teach. AI helps capture the patterns that memory alone cannot hold. Teachers reflect, leaders identify themes, the school adjusts, practice improves, and the loop begins again. That is not surveillance, and it is not automation for its own sake. It is organisational learning.

Why the feedback loop matters

Schools already produce an enormous amount of insight every day. Every lesson contains information about explanation, questioning, modelling, participation, misconception, vocabulary and pupil thinking. Most of it disappears the moment the lesson ends. We rely on memory, instinct and the occasional observation. Those matter, and they are incomplete.

AI changes what can be retained, reflected on and improved. The strategic opportunity is not to generate more content. It is to make professional knowledge more usable. Seen this way, the future of AI in education may not be defined by lesson-planning tools or pupil-facing chatbots at all. It may be defined by whether schools can build trusted systems that help them learn from themselves.

Trust is the strategic constraint

In schools the limiting factor will not be technical capability. It will be trust. Staff will not meaningfully engage with systems they believe are being used to monitor, judge or replace them. Leaders may say the right things, but culture is revealed by use.

Introduce AI as a performance management instrument and it will narrow professional behaviour. Teachers become cautious, perform for the system and avoid risk. The technology collects more data while the school learns less. Introduce it as a private, developmental and teacher-owned tool, and the opposite becomes possible. Staff can experiment, reflect honestly and notice what they would otherwise miss, while leaders support improvement without turning growth into compliance.

This is why schools need an AI strategy, not just AI procurement. The policy, the language and the governance matter as much as the technology.

Using AI well is a discipline, not a default

None of this happens automatically. The difference between AI that flattens thinking and AI that sharpens it lies entirely in how it is used. The same is true at institutional scale. A school that feeds its real questions into these systems, corrects what they get wrong and keeps the human firmly in charge builds something durable. A school that quietly outsources its thinking does not.

The future firm and the future school

Nadella is writing about companies, but the question applies cleanly to schools. What kind of organisation are we becoming. One path leads to growing dependence on generic systems built elsewhere, where staff produce faster outputs while the school’s own knowledge stays scattered, informal and fragile. The other leads to schools as more intelligent institutions that capture what matters, reflect on it responsibly and build professional memory over time.

The second path is harder. It asks for leadership, careful governance, honest conversations with staff and real clarity about what AI is for and what it must never be used for. It is also the path that protects the human centre of education. The aim is not to remove human judgement from schools. It is to strengthen it.

A quiet test for school leaders

A useful test for any AI initiative is simple. Does this help us become more like ourselves at our best, or more dependent on someone else’s generic intelligence? If a tool helps a school preserve its values, reduce unnecessary workload, support teachers and deepen professional learning, it is worth exploring. If it produces more monitoring, more performative paperwork and more distance between leaders and classrooms, it should be challenged.

The best AI systems in education will not be the ones that shout loudest about transformation. They will be the ones that quietly strengthen the learning culture of the school.

This is the thinking behind much of my work with Starlight. The point is not simply that a teacher can record a lesson and receive AI-generated feedback, useful as that is. The deeper opportunity is that a school can build a responsible feedback loop around real classroom practice. Teachers stay in control. Feedback stays developmental and private. Leaders see anonymised, aggregated patterns rather than individual judgements. Professional learning becomes more specific, timely, actionable and regular.

It is also where Nadella's point about owning the loop becomes concrete. We are currently beta testing Deep Field Reports, a bespoke transcript analysis service that lets a school examine its own classroom evidence in real depth and turn it into knowledge it keeps. Over time, the questioning patterns, the explanations that land and the misconceptions that recur no longer vanish at the bell. They accumulate into institutional memory of the kind Nadella describes, built from a school's own practice rather than rented from a generic model. Used well, AI does not replace coaching. It helps coaching scale, and it gives professional judgement more evidence to work with.

The next phase of AI in education will not be won by the schools that adopt the most tools. It will be shaped by the schools that ask better questions. What knowledge do we need to protect. What expertise should we scale. What must stay private. Where can AI reduce workload, and where might it damage trust.

Nadella’s warning is, in the end, a warning about ownership. Organisations that do not build their own learning loops may find the value of their work captured elsewhere. Schools should pay attention, because the future of AI in education is not only about what machines can do. It is about whether schools can become wiser, more reflective and more human because of them.

Adam Sturdee is a senior leader and co-founder of Starlight, the UK’s teacher-first AI-powered transcript-based coaching platform for educators.

His work sits at the intersection of dialogic practice, instructional leadership and responsible AI strategy for schools and trusts.

He recently presented his research on AI-supported coaching at the BERA TEAN Conference 2026: https://www.bera.ac.uk/conference/bera-tean-conference-2026

Schools need more than AI tools. They need learning systems.

Recent Posts

Comments