I read a different kind of book on my holiday this year. Not education, not leadership, not another treatise on change management. It was a book about restaurant kitchens, by an academic called Vaughn Tan, called The Uncertainty Mindset (recommended by my friend, Shonogh Pilgrim). There's a great deal in it I'll come back to across future FRiDEAS pieces I am sure, but one diagram stopped me on the sun lounger long enough to ruin the rest of the afternoon's reading. It maps the different ways new ideas actually arrive in professional cooking, four types, arranged from what a customer directly tastes and sees, through to structures the customer never notices at all. I couldn't stop thinking about it, because I'd recently been on three separate school visits where I watched people mistake one end of that spectrum for the other.

At the first, a multi-academy trust had discovered Gantt charts. Not metaphorically. Someone had built a project timeline in a spreadsheet with coloured bars, and it was presented to me as evidence of a transformed approach to school improvement. It's a scheduling tool. It is over a century old. Henry Gantt died in 1919.

At the second, a group of teachers were extremely excited about an AI tool that writes lesson plans. Type in the topic, out comes a plan. Genuinely useful for saving an evening. Genuinely not a new way of teaching anything, it produces exactly the same worksheet-and-starter structure that's been bolted together in staffrooms for decades, just faster.

At the third, I sat through a presentation on a new marking and feedback policy built entirely around a colour-coding system. Green pen for this, purple for that, a laminated key on every desk. Nobody in the room could tell me what had actually changed about how students used the feedback, only what colour it now arrived in.

None of this is failure exactly. It's all real activity, undertaken by people who mean well and want to improve things. But it sits at exactly the same point on Tan's spectrum, the end where change is loud, visible, and easy to photograph for a newsletter, and where almost nothing underneath has moved.

What Tan Found in Restaurant Kitchens

Tan spent years inside professional kitchens studying how genuinely new ideas get produced, not marketed, produced. What he found is that innovation in food isn't one thing. It's at least four distinct activities, and they sit at different distances from the customer.

Closest to the customer: new dishes, meaning new configurations of already-cooked, already-presented food. This is the flashiest, most visible category and also the shallowest. Most new dishes are incremental, small tweaks to an existing plate, or combinatorial, taking two existing ideas and mashing them together. A handful are what he calls narrative innovation, where the change is really in how the dish is explained to the diner rather than in the dish itself. Same plate, new story.

One step back: new ingredients, the actual things being cooked. Genuinely new ingredients are rare and usually borrowed wholesale from somewhere else, a foraged plant, an industrial byproduct repurposed, so most of what looks like ingredient innovation is really about sourcing, not invention.

Further back still: new cooking methods, the ways of transforming ingredients. This is where kitchens borrow techniques from other trades entirely, sous-vide arrived via food-preservation science, spherification via the chemical industry, and refine or recombine what already exists.

Furthest from the customer, and least visible: new cooking processes, the way an entire kitchen is organised to produce food at all. Nobody eating the dish ever sees this layer directly. It's brigade structure, prep sequencing, how information travels from front of house to back. But it's this layer, Tan argues, that most constrains or enables everything else. A kitchen that can't change how it organises itself will keep producing the same four dishes with different garnishes forever, no matter how creative anyone at the pass tries to be.

The spectrum runs from directly experienced by the customer to less directly experienced. That framing matters more than it first appears. It isn't a hierarchy of importance, it's a hierarchy of visibility. The things customers can see and photograph sit at one end. The things that actually determine what's possible sit at the other, invisible, unphotographed, and usually the last thing anyone tries to change.

We can swap customer for student, or parent, or governor, and you have, almost without alteration, a working diagnostic for school innovation.

‍Four Levels of Educational Innovation

Map Tan's four categories onto education and the parallel is close enough to be uncomfortable. New dishes become new lessons and assessments, the things students directly sit through and are marked on. This is where the Gantt chart trust and the colour-pen school both lived. It's also where the vast majority of what gets called innovation in education actually happens, because it's the layer senior leaders can implement by Monday, photograph for a newsletter, and present as evidence of improvement. Most of it is incremental, a slightly restructured starter activity. Some is combinatorial, bolting a growth mindset slogan onto an unchanged worksheet. A fair amount is narrative, exactly the same lesson, described differently in planning documentation because the description is what gets inspected.

New ingredients become curriculum content, the actual material being taught. Genuinely new content is rarer than schools like to claim. Most curriculum change is really about resequencing existing content, moving photosynthesis from Year 8 to Year 9, or rebranding an existing scheme with a new name to satisfy a subject review. The AI lesson-plan tool sits partly here too. It generates content faster, but it's still drawing on the existing canon of what gets taught and how, just at speed.

New cooking methods become pedagogy, the actual method of teaching. This is where borrowing from adjacent fields happens, retrieval practice from cognitive psychology labs, cold-call questioning from Direct Instruction research, structured talk from oracy research. Real, and harder to fake than a new lesson, because it requires teachers to change habitual classroom behaviour rather than swap a resource.

New cooking processes become school structures, timetabling, staffing models, how information and decisions actually move through a building. This is the layer almost nobody touches, because it's invisible to students, invisible to parents, and genuinely difficult. It means changing who has authority over what, how staff time is allocated, how a curriculum is actually built rather than bought. A trust's Gantt chart lives, in principle, at this level, project management is a process tool. But adopting the tool isn't the same as changing the process it's meant to serve. You can run a beautifully colour-coded Gantt chart tracking a school improvement plan that changes nothing about how decisions are actually made in that building. The artefact of process innovation without the substance of it.

This is the trap. Schools consistently mistake activity at the shallow end, new lessons, rebranded content, adopted tools, for change at the deep end, pedagogy, structure. The AI tool felt revolutionary to the teachers using it because it changed their Tuesday evening. It changed nothing about how the students in front of them on Wednesday morning would be taught, what they'd be taught, or how the school around them was organised to support that teaching. New ingredient, arguably but definitely not new method or new process.

‍Why Schools Stay at the Shallow End

There's a reason this happens everywhere, and it isn't laziness or a lack of imagination. It's incentive structure. Shallow-end innovation is fast, visible, and low-risk. A new marking scheme can be rolled out this half-term, photographed for the school website, and reported to governors as evidence of progress. Deep-end innovation, changing how a school allocates staff time, or how curriculum decisions actually get made, takes years, is invisible to most stakeholders, and can go wrong in ways that are politically expensive for whoever championed it.

Inspection windows reward the same asymmetry. A visiting inspector spends, at most, a couple of days in a building. They can see a lesson. They can see a marking policy applied to a set of books. They cannot see whether the underlying decision-making structure of a school has genuinely changed, because that isn't visible in a two-day snapshot, it's visible over years. So schools, entirely rationally, invest disproportionately in the layer that gets inspected.

There's also a genuine cognitive bias at work, visibility gets confused with importance. A new dish is tangible, you can taste it, photograph it, put it on a menu. A new kitchen process is abstract until something goes wrong. The same is true of a lesson versus a staffing structure. It is far easier for a senior leader to feel they've achieved something when they can point at a laminated colour key than when they've spent eighteen months quietly renegotiating who gets a free period on a Wednesday.

None of this is cynicism. The MAT with the Gantt chart genuinely believed they'd found something significant, and in a narrow sense they had, project visibility is a real improvement over no visibility at all. The teachers with the AI tool genuinely saved themselves hours. These are real, if modest, gains. The danger isn't that shallow-end innovation is worthless. It's that mistaking it for deep-end innovation lets a school believe it has changed fundamentally when it has, in Tan's terms, produced a new dish from an unchanged kitchen.

Donella Meadows and the Problem with Easy Levers

The systems theorist Donella Meadows made almost exactly this argument about complex systems generally, decades before Tan wrote about restaurant kitchens specifically. In her essay on leverage points, places where a small intervention in a system produces a large effect, she ranked interventions from the shallowest, adjusting numbers such as budgets, targets, or class sizes, to the deepest, changing the paradigm the whole system is built on, the goal it actually serves.

Her list runs, roughly, from tweaking numerical parameters at the bottom, through changing the structure of information flows, then the rules of the system, then who has power over those rules, and finally, at the very top, the paradigm the system exists to serve. Almost everything schools do sits at the bottom of that list. New marking colours are a parameter change. A new timetable structure is closer to the middle, an information-flow and rule change. Almost nothing touches the top, what a school believes education is actually for, because that's the layer that determines everything beneath it, and it's also the layer people are least willing to disturb.

"The higher the leverage point, the more the system will resist changing it" Donella Meadows

That resistance isn't incidental. It's the system defending itself, and it explains precisely why the Gantt chart trust found it easier to adopt a new spreadsheet than to ask what its improvement plan was actually trying to achieve.

There's a genuine irony worth sitting with here. The shallow interventions, the ones schools reach for most often, are also, on Meadows's own analysis, the ones least likely to produce lasting change. You can adjust a budget line or a marking colour indefinitely and the system keeps behaving exactly as it did before, because none of those adjustments touch the goal the system is actually organised around. If a school's real, functioning goal is to look improved by the next inspection, a new colour-coded policy serves that goal perfectly. It just doesn't serve the goal most staff think they're working towards.

Bernstein and the Curriculum Nobody Sees

The sociologist Basil Bernstein got at the same problem from inside the classroom rather than the system as a whole. He distinguished between what he called visible and invisible pedagogy, the explicit rules of a classroom, what's assessed, how it's marked, what's on the wall, versus the implicit, unstated rules that actually govern who succeeds, whose way of talking is rewarded, whose background assumptions are treated as normal, what good work is quietly understood to look like. His argument, reduced considerably, was that changing the visible layer, a new marking scheme, a new assessment format, does almost nothing to the invisible layer beneath it, and it's the invisible layer that does most of the sorting. A school can introduce the most progressive-looking visible pedagogy imaginable, open-ended tasks, no formal grading, and still reproduce exactly the same pattern of who succeeds, because the invisible rules about what counts as good thinking haven't moved at all.

This maps directly onto the colour-pen school. The visible layer changed completely, every book in the building now has purple and green annotations in the right places. The invisible layer, what actually counts as good feedback, whether students have any real opportunity to act on it, whether teachers have the time and training to give it meaningfully, was never addressed. The policy was, in Bernstein's terms, a change to the visible pedagogy with the invisible pedagogy left entirely intact.

This is worth sitting with because it cuts against a comforting assumption, that visible change is at least evidence of good intent, a step in the right direction. Bernstein's point is sharper than that. Visible change can function as camouflage for the invisible layer staying exactly as it was, and can even make that layer harder to see, because everyone in the building is now looking at the new colour key rather than asking the harder question underneath it.

What Genuine Innovation Actually Requires

None of this is an argument against Gantt charts, AI tools, or coloured pens. Shallow-end innovation isn't fake, it's just partial, and pretending otherwise is what causes the trouble. Genuine innovation in a school means being honest about which layer you're actually working on, and not mistaking movement at one layer for movement at another. If you've changed a marking policy, say so, and don't claim it's transformed pedagogy unless you can point to what's different in how a teacher actually responds to a piece of work. If you've adopted a project-management tool, credit it for what it does, visibility, not for what it doesn't, changed decision-making.

It also means occasionally being willing to work at the level that resists you most, Meadows's top of the list, the goal the school actually serves, Bernstein's invisible pedagogy, the unstated rules about who gets to succeed. That work is slow, mostly invisible to parents and inspectors, and offers very little to photograph. It is also the only work that changes what a school actually is, rather than what it looks like. The MAT with the Gantt chart wasn't wrong to want visibility over its improvement work. It was wrong to believe visibility was the improvement. The teachers with the AI tool weren't wrong to want their evenings back. They were wrong to call it new teaching. The colour-pen policy wasn't wrong to want better feedback. It never got near asking what better feedback would actually require.

Tan's kitchens teach the same lesson food already knew. A new dish is real. It is also, on its own, the shallowest kind of change a kitchen can make. Recognising that doesn't devalue the dish. It just stops you mistaking it for a new kitchen.

‍Key Takeaways

1. Most school "innovation" happens at the shallowest, most visible layer - this is the layer that's easiest to launch and easiest to photograph
2. A new lesson or resource is real change but it's the easiest kind to make, which is exactly why it gets mistaken for something deeper
3. The deepest change, how a school is actually structured and how decisions get made, is invisible to parents and inspectors, which is exactly why it gets left alone
4. The changes with the most power to shift a school are also the ones that meet the most resistance, so leaders default to the ones that meet the least
5. Rewriting a policy doesn't automatically rewrite the unwritten rules about who actually succeeds, and those unwritten rules are usually what's really running the show

I'll come back to Tan's book again. There's a section on uncertainty tolerance in professional kitchens that has direct implications for how we think about teacher workload, and another on knowledge transfer between chefs that I think says something uncomfortable about how badly schools share good practice. But this felt like the place to start, because it's the diagnostic every leadership team should run before announcing their next transformation. Ask which layer you're actually changing. Ask whether the layer beneath it has moved at all. A new dish can be wonderful. It can also be exactly what a kitchen serves when it doesn't want to change anything else about how it cooks. Before calling anything a transformation, name the layer you actually changed, that honesty is the difference between decoration and progress.