Some thoughts on the future of edtech...

Let's have an honest conversation about something the software industry would rather you not think about. A lot of the software you are currently paying for monthly — the dashboards, the productivity tools, the "AI-powered" lesson planners with a suspiciously basic feature set — is about to become structurally unnecessary. Not because it's bad. But because building a working alternative is now, genuinely, a five-click problem.

Platforms like Google AI Studio and a growing roster of competitors have quietly crossed a threshold. You can describe what you want, watch it get built in front of you, and share a live URL with your students before your coffee goes cold. No developer. No agency. No six-month roadmap. No per-seat licensing. Just: here's a thing, here's a link, go use it.

This changes quite a lot. Let's talk about what.

The classroom gets its own software department

Consider what this means for teachers specifically. Right now, a teacher who wants a custom interactive exercise — say, something that shows how French verbs are synthetic and derived from Latin root forms, with animations and a little quiz at the end — has two realistic options: find something that almost fits on the App Store, or build a makeshift version in a PowerPoint and quietly despair.

In the very near future, that same teacher spends fifteen minutes describing what they want, refines it with a few prompts, and has a live web app they can share with their class that afternoon. Next week, they make a different one. For a different lesson. About the gradual Roman advance into Gaul, complete with a dynamic map that moves with the narrative. Or a virtual gravity experiment where students can adjust mass and height and watch the physics play out.

"Bespoke" used to mean expensive. It's about to mean Tuesday afternoon.

The implications for EdTech are significant. The model of paying £8 per student per month for a platform that gives you access to a library of content someone else made, for a curriculum that isn't quite yours, for students whose context the platform has never met — that model has a credibility problem it hasn't fully reckoned with yet.

What actually survives, and why

To be fair — and this article is trying to be — not all software is equally vulnerable to this shift. There is an entire category of software that is less about the code itself and more about everything wrapped around it: the compliance infrastructure, the legal agreements, the payment processing certifications, the cybersecurity architecture that took years and serious engineering budgets to build.

Stripe is not going anywhere. Cybersecurity platforms built on proprietary threat intelligence are not going anywhere. Streaming services with their expensive content licences and their deeply unglamorous but deeply real DRM systems are not going anywhere.

The honest rule of thumb: if the value of a software product lives primarily in the code — in the interface, the UX, the feature set — it is vulnerable. If the value lives in the compliance certifications, the legal agreements, the exclusive data, or the costly infrastructure underneath — it probably isn't. Most EdTech sits very firmly in the first category.

What is going away — or at least, what is going to have a very difficult few years — is pure-software products whose value proposition is essentially "we built this so you don't have to." That bargain made complete sense in 2015. It makes considerably less sense when "building it yourself" takes an afternoon and costs nothing beyond your AI subscription.

Not pay to use. Pay to build.

The payment model shifts too, and this is perhaps the most interesting part. People will increasingly pay a monthly fee to access the platform that lets them build and host software — and stop paying for the software itself. The tool to make the tool becomes the product. OpenAI, Google, Anthropic and others are, in a sense, becoming the new enterprise software vendors. Except instead of selling you a finished thing, they're selling you a workshop.

The numbers already confirm this shift is well underway — and the pace of it is genuinely difficult to absorb.

Anthropic went from $1 billion in annualised revenue in December 2024 to $9 billion by the end of 2025 and over $30 billion by April 2026. That is not a growth curve. That is a vertical line. SaaStr, tracking B2B software for over a decade, called it flatly: "There is simply no precedent for this in B2B software. Not Slack, not Zoom, not Snowflake. Nothing." The majority of that revenue comes not from consumer subscriptions but from enterprise and developer API usage — businesses paying to access the tools to build their own things.

Claude Code — Anthropic's AI coding tool — went from zero to $2.5 billion in annualised revenue in roughly nine months. For context, it is now reportedly authoring around 4% of all public GitHub commits, with projections of 20%+ by end of 2026. Developers are not just using it to write code faster. They are using it to build software that previously required a team.

Google tells a similar story. Gemini's API processed 85 billion requests in January 2026 alone — a 142% increase from March 2025. Google Cloud grew 34% year-on-year, driven primarily by enterprise demand for Gemini-powered services. Alphabet crossed $400 billion in annual revenue for the first time, and its CFO has been explicit that AI infrastructure is the engine. Over 2.4 million developers are now actively building on the Gemini API, up 118% in less than a year.

The pattern is consistent across both companies: the people paying are not end users consuming a finished product. They are builders. Teachers, developers, entrepreneurs, and enterprises paying for access to the workshop so they can make exactly what they need, exactly when they need it.

This is not unlike what happened to music production. A generation ago, access to a professional recording studio was the bottleneck. Then GarageBand happened, then Logic, then cloud distribution, and suddenly the bottleneck moved. The studios that survived either went ultra-premium or pivoted to services. The middle quietly hollowed out.

Software is about to experience something structurally similar. The ultra-premium end — complex regulated products with serious infrastructure requirements — stays. The genuinely creative and customised end — individual teachers, small teams, solo practitioners — gets handed the tools to do it themselves. The middle, the "good enough for most people" SaaS products charging £15/month for features that an AI can now replicate in twenty minutes, is going to have to find a more compelling reason to exist.

What this means if you're building in EdTech right now

The EdTech products worth building in the next five years are not the ones that replace teacher judgement with a library of pre-made content. They are the ones that give teachers genuine leverage: infrastructure that works in difficult environments, workflows that save real time, tools that handle the parts of teaching that genuinely don't scale — feedback, differentiation, tracking — rather than the parts that AI will happily generate for free on a Tuesday afternoon.

The teachers who understand this shift will be remarkable. They'll be running their classrooms like small software studios, generating bespoke tools for individual lessons, experimenting with formats that didn't exist last semester, and discarding them without guilt because the marginal cost of the next idea is approximately zero.

That is, frankly, a very exciting classroom to be in.

The question is no longer "can we afford to build this?" It's "can we afford not to imagine it?"

The software that survives this era will be the software that earns its place — not through switching costs and annual contracts, but because it does something genuinely difficult that the five-click alternative cannot. Everything else is, with affection and a little sarcasm, on borrowed time.