Half of Edtech Exists to Fix the Other Half

I've spent years building education technology. Here's the uncomfortable truth I keep returning to: roughly half of edtech exists to fix problems created by other edtech.

Watch the pattern:

A school buys a Learning Management System. The LMS is clunky, so they buy integration tools. The reporting is unreadable, so they add analytics dashboards. Messaging is atrocious, so they layer on a communication platform. Assignments are basic, so they bolt on assessment plugins. Collaboration is non-existent, so here comes Padlet. The student experience is hostile, so they wrap it all in a portal.

Then add cohort mapping. Group formation tools. Engagement tracking. Attendance automation. Plagiarism detection. Accessibility checkers for the inaccessible content. Writing assistants. Citation managers.

The average US school district now accesses 2,982 distinct edtech tools annually, a 9% year-on-year increase and nearly triple what it was in 2017-18 (Instructure/LearnPlatform, 2025).

This isn't an ecosystem. It's a Rube Goldberg machine where everyone's selling duct tape.

Sold to Administrators, Suffered by Teachers

Here's the dirty secret: the people buying edtech are almost never the people using it.

District IT and procurement teams evaluate software against compliance checklists, security certifications, and feature comparison spreadsheets. Teachers and students, the humans who must actually navigate these tools daily, rarely get meaningful input.

Digital Promise's edtech procurement research found that "many school districts continue to use historical and antiquated procurement practices based upon older textbook adoption models." Products are purchased "without properly vetting their effectiveness and usefulness with end-users."

So we get software optimised for the RFP, not the classroom. Tools that impress in 30-minute demos but drive teachers to despair by Week 3.

Teachers are already stretched past breaking. EdWeek Research found they work 54-hour weeks, with vast chunks devoured by administrative tasks. Sixteen percent left the profession or changed schools in 2020-21 alone.

They don't need more software. They need less admin.

The Wrong Friction Is Being Removed

Meanwhile, the new wave of "AI-powered" edtech commits a different sin: removing friction from learning itself.

It sounds progressive. It's pedagogically dangerous.

Learning research is clear: productive struggle is essential for deep, lasting learning. When students wrestle with challenging material, when they hit walls and push through, they develop critical thinking, perseverance, and genuine understanding. Remove that struggle, and you get shallow, rapidly forgotten knowledge.

Bellwether's 2025 study found students using ChatGPT for homework showed improved short-term performance but worse long-term retention. Researchers coined it "metacognitive laziness": students outsourcing thinking rather than doing the cognitive work.

Another study tracked high schoolers given ChatGPT access. Students who used it as a "crutch" during practice performed worse when working independently. The AI helped them appear competent while making them less capable.

Here's what makes this insidious: it feels like learning. The homework gets done. The grades look fine. But the brain work that produces actual knowledge? Skipped entirely.

WestEd's "Friction by Design" framework captures this perfectly: there's productive friction (the effort and struggle that drive growth) and unproductive friction (barriers that just waste time). Most edtech obliterates both indiscriminately, and markets it as innovation.

The Reckoning Is Here

The market has noticed. Between January 2023 and December 2024, edtech saw 40,000+ layoffs across 230 separate events, an 18% reduction in headcount. Over 2,100 startups shuttered. The easy funding has evaporated.

But this isn't just a correction. It's the beginning of obsolescence.

AI Agents Don't Need Your Middleware

The entire edtech stack was built for a 2016 reality: software that helps humans perform tasks. Click this. Fill that. Navigate here. Copy data there. Integrate System A with System B through a third system built specifically to connect them.

AI agents don't work that way.

They don't need a polished student portal. They don't need a "better assignments" plugin. They don't need a dashboard to interpret another dashboard. They access the underlying data and just... do the work.

Gallup (2025) found teachers using AI tools save an estimated 5.9 hours per week. Not through prettier interfaces, but by having AI perform work that previously required wrestling with multiple systems.

Bain projects AI agents will handle 60% of routine SaaS tasks by end of 2025. Forbes describes an emerging "agentic layer": intelligent systems transforming software from interfaces-to-navigate into collaborators-that-act.

When an AI agent can directly create differentiated assignments, analyse student performance patterns, draft parent communications, and adapt instruction in real-time, what exactly does the middleware layer contribute?

The answer is becoming clearer by the month: nothing.

What Survives

The parasitic edtech layer (the integrations, the dashboard enhancers, the "better than native" plugins) is living on borrowed time. Its entire value proposition depends on other software being bad.

The tools that survive will be ones that:

1. Demonstrably improve learning outcomes, not just paper over UX failures
2. Preserve productive struggle while eliminating unproductive friction
3. Respect teachers as professionals rather than treating them as data-entry clerks
4. Prepare for agentic futures instead of optimising for 2016 procurement cycles

Everything else is technical debt waiting to be written off.

A question for educators and edtech builders: What's the most redundant tool in your stack, the one that exists purely because another tool doesn't do its job?

I suspect that answer tells us more about our industry than we'd like to admit.