Because it matters (for reasons I’ll get into another time), I need to say that this was all written by me. AI helped with research in a couple of places as noted.

The impact of AI on education could not be more universal. It fundamentally disrupts the purpose of education, how we achieve that purpose, and how we know we’ve achieved that purpose. Sounds like pretty much everything, no?

• The purpose of education. Of course our job is to prepare children to enter the world with confidence in their competence. Historically that has translated to a purpose of cataloguing the skills and knowledge that the world values and constructing a pedagogical ladder to reach them. But AI is imposing a fundamental shift in terms of what the world needs from humans. For instance, the acts of constructing, transforming, and analyzing information are now, for the most part, readily performed by AI. We can’t see the end of that shift yet. But it has already moved away from many of the skills that we invested so much educational development in. Knowledge work, long seen as an apex skill, the culmination of a progression starting in the Agricultural Revolution, is now in the wheelhouse of AI. What world are we supposed to prepare children for and how can they be confident in meeting it if we aren’t confident in even knowing what it is?
• How we achieve it. Ignoring the utility of AI in teaching and learning is a catastrophic error in the world we’ve already arrived in. If, ten years ago, I had waved a magic wand and provided every student on the planet with an expert tutor in every subject, around the clock, free of charge, all of these tutors with infinite patience and good humour, I would either have been hailed as the greatest fairy godfather since Cinderella or tied to the stake as a witch. But no one would have turned those tutors away. That would have represented the greatest boon to a child’s development since the invention of the printing press. But that of course is exactly what we have now. Yet some people would turn it down because those tutors in their electronic form will, if not configured appropriately, also just do the student’s homework for them. We can – we must – handle that.
• How we know we’ve achieved it. Knowing what competence a child has achieved is the basic marker of what that child can do next in the world – what higher education path they can handle, or what job they are qualified – and also, in aggregate, the basic marker of how good a school is at its job. Livelihoods hinge on third decimal place differences in carefully curated statistics of those markers. Entire industries are constructed on the methodologies of assessment. But these are now disrupted all the way up and down the chain. At the point of application, there is academic integrity – if a student can use AI to complete an assignment without our being able to tell, their assessments are wrong and other students’ incentives are destroyed. If they can use AI in the process of completing an assignment that results in the creation of a product that passes assessment, but they did not experience the learning that we wanted and expected the process to imbue, we have failed. And of course if they’ve learned something that’s no longer of utility in the world, the assignment was pointless to begin with.

As damning as all of these disruptions are, they are not the biggest problem at the moment.

The biggest problem is that you already know these things, and yet education isn’t changing the way you know it should. Because of how many of your colleagues who don’t and won’t read this. You, as a reader of this kind of article, are still exceptional in the field. We need the others to become the exceptions.

We need a critical mass.

What critical mass does a revolution need? And this is where AI comes in handy to compensate for my meagre knowledge of history. Erica Chenoweth has shown that only 3.5% of a country’s population is needed in active support of a revolution for it to succeed.

If only it were that easy to bring about a revolution in education.

The problem is more like corporate cultural change. Again, AI to the rescue. Research suggests that the participation of 20-30% of a workforce can bring about permanent change, and as few as 3-5% of that workforce can, if aligned, bring about that.

Is critical mass the 3-5% or the 20-30%?

I think there’s another problem we have to surmount first. Education is nearly as hierarchical as the Catholic Church. Structural change is not something it undertakes easily. As much as it may look otherwise to parents who wonder why schools have to continually invent new ways of teaching math that void their attempts to help their children with their homework, those changes aren’t undertaken without a lot of air cover. Lengthy longitudinal studies in university schools of education, consolidated through meta studies, picked up in government departments or ministries, pushed down in recommendations from districts, so that school leaders can feel that what they are doing is safe for students, has copious evidence of that safety, and is aligned with what other schools are doing so that one school’s students don’t come out looking like they’ve graduated from Mars.

AI has driven a stake through the heart of that model.

Change is now simply happening far too fast for that model to help schools adapt fast enough. Yet, understandably, school leaders are still looking to that system, hoping that it will any day now emit the direction they desperately need.

By the time a direction comes out of that system, AI will have evolved again and render it obsolete.

The hierarchy of change has become inverted. The point of innovation is now the classroom. That’s where change is being experienced and engaged with daily. Individual schools, and small associations of schools, are in the driver’s seat of innovation in education, like it or not.

I see places that have realized this. But there are three stages to managing this inversion, and they have only completed the first two:

• The first stage for a school or its teachers is realizing that the inversion of the innovation hierarchy has taken place.
• The second stage is realizing that this means that being in the driver’s seat means taking hold of the steering wheel; that they must act to instigate change.
• The third stage is creating the power structures that enable that change to matter.

The symptoms of being stuck at stage two look like holding guest lectures on professional development days to learn about AI. (I’ve delivered many of these personally.) As a tool for the 3-5% to build the 20-30%, that’s a great start. And then that develops into seminars for wider audiences.

But the established hierarchy comes with structures of authority that drive workflows (and let’s not forget funding) that are the fuel for the engine of change. That engine is stalled. The inverted hierarchy is toying with building an engine but it has no fuel. We need mechanisms that result in changes in school curricula, in assessment standards, in redesigning how time is spent in the classroom. Now, not in five years.

How an inverted hierarchy could effectively manage change is the big question. I realize that what I’m saying sounds like it’s arguing for creating a new institution, but the language alone is an albatross. Doesn’t the dictionary say something like “institutionalization, n., see ossification”?

But I submit this is our current challenge. To realize that we have to stop looking for direction to come from a system that is incapable of delivering it in time, and that that means we have to both take charge and build a bottom-up system that can rapidly design and implement educational change at the point of delivery at the scale of multiple schools connected by geography or philosophy.

No one said this would be easy. But the thing that gives hope that it’s possible is embracing the creative use of state-of-the art AI, through, for example, agentic workflows that accelerate business process development.

What stage are you, your school, and your peer schools at?