I understand how confronting it can feel when you’re facing an uncertain future that you’re not quite ready for.

On a personal level, it’s understandably unsettling, but from a school leadership perspective, it becomes even more pressing.

This is the core challenge highlighted by a recent Australian survey of parents. Only 38% felt that their child’s school is doing a good job of preparing students for a future working world influenced by artificial intelligence. Nearly half of parents are worried that AI could significantly affect job opportunities for their children’s generation.

For those who’ve been around the block once or twice, this response shouldn’t be too alarming. At first glance, it seems obvious. Schools need to move faster.

If only we could introduce more AI, better staff training, opportunities for innovation and experimentation and then get that into classrooms – as soon as possible.

Perhaps.

But history is littered with organisations that correctly identified a technological revolution, moved early and still managed to lose.

Which raises a different strategic question for schools.

Does moving FIRST create an advantage, or merely make you the first to pay for everyone else’s learning?

Being right about the future is not enough

At the height of the dot-com boom, Webvan looked like the future of grocery shopping.

Its proposition now seems almost mundane: order groceries online and have them delivered to your door.

In 1999, it looked revolutionary.

Investors agreed, and faster than money can be printed, Webvan raised around US$1.2 billion. It built sophisticated automated warehouses and embarked on an ambitious expansion strategy across the United States. Less than two years later, it was bankrupt.

The irony is that Webvan was not wrong about the future.

We do buy groceries online.

Where they erred was in how quickly customers would change their behaviour, how much infrastructure would be required, how the economics would work, and how rapidly it should scale. Webvan committed enormous resources before enough of those assumptions had been tested.

Other supermarkets learned from the experiment. Coles and Woolworths, for example, approached online grocery differently, using much of their existing store infrastructure rather than building an entirely new distribution network from scratch.

The future Webvan imagined arrived.

There is a lesson here for schools racing to become AI pioneers.

AI in learning: use isn't the same as value

The parent survey is particularly interesting because parents weren't simply asking schools to expose students to more AI.

More than 75% said that capabilities, including critical and creative thinking, resilience, character, and problem-solving, should be prioritised either equally with, or ahead of, academic results. Some respondents were explicitly concerned that over-reliance on generative AI could undermine children's ability to think independently.

This highlights the strategic problem dramatically.

If the objective is simply AI adoption, measuring progress is easy. Basic metrics emphasise these:  licences, users, prompts, the number of subjects incorporating AI.

But none of those measures tells us whether a student is becoming a better thinker.

There is emerging evidence that AI can support learning. A 2026 randomised experiment involving university students found that access to generative AI improved immediate knowledge scores, with those gains persisting when students were tested again one week later. More interestingly, the researchers found different outcomes depending on how the technology was used. Students who used AI to augment their thinking, such as asking it to explain concepts, experienced more enduring benefits than those who primarily used it to automate writing.

It is one experiment involving university students, which of course is not proof of what will happen across Australian schools. But the distinction is worth noticing.

Perhaps the strategic question isn't whether students should use AI.

Perhaps it is what intellectual work we still expect students to do when they use it.

That feels much closer to the intent of Australia's national framework for generative AI in schools, which places teaching and learning alongside human wellbeing, transparency, fairness, accountability, privacy and safety. The framework recognises the potential of generative AI while also warning of risks to assessment integrity, privacy, accuracy and bias.

Being AI-ready therefore cannot simply mean giving students access to the technology.

A school could have AI embedded everywhere and still be poorly prepared for an AI-shaped future.

The second AI opportunity sits behind the classroom

There is another side of this conversation that receives less attention.

AI may change what students learn and how they learn it, but it may also change how schools themselves operate.

Consider a raft of administrative tasks that schools require: lesson planning, resource creation, reports, parent communications, policies, meeting summaries, yada yada…

Here, the value proposition is somewhat easier to measure because time can be calculated.

A randomised trial conducted by the Education Endowment Foundation involving 259 teachers across 68 English secondary schools examined the use of ChatGPT for Year 7 and 8 science lesson preparation. Teachers using ChatGPT spent an average of 56.2 minutes per week preparing lessons and resources, compared with 81.5 minutes for teachers not using generative AI.

That represents a 31% reduction in planning time, without the expert panel detecting a reduction in the quality of the resources produced.

If that saving can be replicated across subjects and across the many administrative tasks that consume teachers' and leaders' weeks, AI could create something schools desperately need.

Time.

And schools appear to be moving in that direction. The UK Department for Education's Technology in Schools Survey 2024 to 2025 found that 44% of teachers were already using generative AI for school activities, most commonly for lesson planning. More revealingly, school leaders were considerably more likely to be planning investment in AI tools for teachers than for pupils: 58% compared with 20%.

Perhaps the less glamorous AI revolution in schools will happen in the staffroom before it happens in the classroom.

But saved minutes don't automatically become strategic value

This is where I become cautious.

Businesses are discovering something similar.

A major 2026 NBER study surveyed nearly 6,000 senior executives across the United States, United Kingdom, Germany and Australia. AI adoption was already widespread, with 69% of firms reporting active use.

Yet 89% reported no impact on labour productivity over the previous three years.

Among the minority that did report productivity improvements, the estimated average effect across all firms was just 0.29%. Curiously, those same executives remain optimistic, predicting an average productivity improvement of 1.4% over the next three years.

So we have high adoption, small realised organisation-wide gains and considerable confidence that the larger benefits are still coming.

Schools should pay attention to that gap.

Saving a teacher 25 minutes of lesson preparation is useful. But what happens to the 25 minutes?

Does it disappear into another administrative task? Or does it allow richer feedback to students? Could it reduce work taken home? And perhaps increase time spent collaborating with colleagues?

Or does the school pay for another platform while workload remains stubbornly unchanged?

The technology might be working perfectly while the organisation captures almost none of its value.

So when does AI become a competitive advantage?

Probably not when a school buys it.

The technology itself is unlikely to provide much of a moat. If one school can subscribe to an AI platform this month, its competitors can probably subscribe to it next month.

Competitive advantage begins to look more plausible when something difficult to copy develops around the technology.

This is where innovation needs to become strategy, rather than a collection of disconnected experiments. The question is not simply what a school can do with AI, but how those innovations create value that supports its broader strategic position.

For students, that might be a distinctive pedagogical approach in which AI strengthens rather than substitutes for thinking. Teachers become skilled at knowing when students should use AI, when they shouldn't, and how assessment can reveal genuine understanding.

Organisationally, it might mean redesigning work rather than simply automating fragments of existing processes. If AI saves staff time, the school deliberately decides where that time should be reinvested and then measures whether the benefit actually appears.

Over time, the advantage isn't the AI.

It is the capability that the school develops by learning to use it well.

Those things take considerably longer for another school to copy.

Learn first. Scale second.

Webvan's story isn't an argument against innovation.

Quite the opposite.

Its mistake was treating a plausible future as though all the assumptions required to reach that future had already been proven.

Schools have an opportunity to take a different path with AI.

Experiment. Measure. Learn. Expand what works. Abandon what doesn't.

For student-facing AI, we should ask whether learning, independence, and intellectual capability are improving.

For staff-facing AI, we should ask whether workload, cost, quality, and organisational capacity are actually changing.

And perhaps we should become slightly less impressed by adoption metrics.

The pioneering school may not ultimately be the school with the most AI.

It may be the school that learns fastest where AI creates genuine educational and organisational value, where it doesn't, and has the confidence to tell the difference.

Because being early is not a strategy.

Turning what you learn by being early into something others struggle to replicate might be.