Schools are Legacy Systems

Every sector eventually faces the same question: do we modernise, or do we keep patching what we've got and hope it holds?

Banking faced it. Healthcare faced it. Retail, manufacturing, government. They all hit a point where the systems they'd built decades ago stopped being reliable foundations and started being the thing holding them back. The people who understood those systems retired. The documentation was never written. And everyone left behind developed a quiet, unspoken agreement: don't touch it, it works, probably.

Education hasn't had its modernisation moment yet. But it's coming. And the weird thing is, the best roadmap I've found for how to think about it didn't come from an education report or a policy paper.

It came from a playbook about programming.

Stay with me.

The COBOL problem

There's a programming language called COBOL. You've never heard of it. But it handles the majority of banking transactions worldwide. Hundreds of billions of lines of it, running right now, while you're reading this.

Almost nobody left knows how it works.

The people who built these systems retired. The code's been patched for decades. The knowledge walked out the door with the people who had it. Sound like any school you know?

In February 2026, Anthropic published a blog post about using AI to modernise these systems. The argument: understanding legacy code used to cost more than rewriting it. So nobody bothered. Everyone just kept patching.

AI changes that maths. Completely.

Share prices tumbled overnight. Consulting firms that had built empires around the difficulty of modernisation saw billions wiped off. The game changed in a day.

But here's the bit that matters to you. Alongside that blog post, Anthropic released the Code Modernisation Playbook. And buried in it are ideas that have nothing to do with programming and everything to do with your school.

You're running a legacy system

The timetables. The curriculum frameworks. The way we batch children by age and move them through content on a fixed timeline regardless of whether they're ready. The whole thing.

It was built for a world that needed standardisation. And it worked. For ages. The core logic is still sound. Children need to learn to read, write, think, collaborate. That hasn't changed.

But the delivery system around that logic? That's a 1960s mainframe trying to run a 2026 economy.

You feel it. We're preparing children for a world that doesn't exist anymore. You know it. Your staff know it. Your pupils definitely know it.

The trap

The playbook says real modernisation isn't about throwing away decades of refined logic. It's about transforming how that logic lives in your infrastructure.

Read "logic" as "learning." Read "infrastructure" as "how we run schools."

Tell me that doesn't land.

McKinsey says developers spend over 17 hours a week on maintenance and technical debt instead of building new things. Protiviti found nearly 70% of organisations say technical debt is the primary thing blocking innovation.

Now think about your teachers. How much of their week is compliance paperwork, marking busywork, data entry into systems that don't talk to each other, meetings about meetings? How much of your day is spent keeping the machine running instead of asking whether the machine still works?

The playbook calls this a "vicious cycle where innovation becomes nearly impossible." In schools, we call it Tuesday.

Here's the part that should properly worry you. When legacy systems go unmodernised, organisations develop a culture of fear around touching anything that appears to be working. "Working" quietly starts to mean "we're afraid to check."

That's every staffroom conversation about assessment policy. Ever.

What you actually do

Stop reading about change and start making one. The playbook says start small. High-visibility, low-risk. Prove the concept. Build confidence. Expand.

Fix one broken thing where everyone can see it. Your reporting system. Your intervention scheduling. The way NQTs get onboarded with zero documentation, just like those COBOL systems where all the knowledge lives in one person's head. Pick it. Fix it. Use AI to help. Show people it's possible.

Separate the mission from the machinery. The mission (helping young people think, create, collaborate, navigate complexity) is timeless. The machinery we wrapped around it needs updating. Protect the mission. Modernise the machinery. Stop muddling the two.

Your pupils are already there. 92% of university students were using AI tools by 2025. In schools, 85% of teachers and 86% of students used AI last academic year. Your pupils are living in the modern architecture already. Meet them there. Stop pretending the old system matches their world. It doesn't.

Use AI as a thought partner. The best results come when you use AI to think through problems, not just generate output. Anthropic built Claude for Education around this. Its Learning mode asks "How would you approach this problem?" instead of giving the answer. That's good pedagogy. That's your model.

Document everything. Now. Institutional knowledge vanishes when people leave. Start capturing what your best teachers know. Use AI to help build it. Don't let your school become a system that only works because one person understands it. You know exactly who I'm talking about.

So. About that modernisation moment.

Remember the question at the top? Every sector faces it. Modernise or keep patching?

The tech world just learned that the cost of understanding old systems has dropped off a cliff. AI can do in days what took consultants months. Banking is modernising. Healthcare is modernising. The companies that made money from the difficulty of change are scrambling because the difficulty just evaporated.

Education's turn is here. Right now. The tools exist to personalise learning, cut teacher busywork, surface what pupils actually know, and free up time for the human work no algorithm can touch.

The question isn't whether schools will modernise. They will. The question is whether you'll lead it or watch it happen around you while you're busy maintaining a system that stopped working years ago.

COBOL still runs the cash machines. Nobody's building new systems in COBOL.

What are you building?