AI Is Making Edtech Pricier—and In-House Builds Plausible Again
Educational technology has never been my favorite slice of capitalism. Universities and other large bureaucracies are slow to change, as we’ve discussed in other areas. That inertia creates ideal conditions for vendor lock-in: once a contract is in place, the cost of switching—financial and organizational—gets enormous. From there, profitability tends to follow two paths: (1) ratchet prices upward year over year, or (2) save costs and let the product stagnate. Most of us have used a campus system that still feels like the 1990s; this is one reason why. The end result is familiar: we pay a lot for tools that aren’t very good. These companies have optimized for that dynamic.
Enter artificial intelligence.
There are two ways AI is changing this relationship, one predictable and one completely novel.
The predictable change: the “AI tax”
Vendors in all areas are stapling AI onto existing products and charging more. We were early to criticize Turnitin’s AI-detection tool, which was enabled without notice. Now that add-on costs more. It’s easy to imagine it becoming a default feature folded into a pricier bundle. More recently, OpenAI announced a partnership with Instructure (Canvas). Across the sector, the pattern repeats: attach an “AI feature” or a ChatGPT wrapper to the same product, then raise the price. That’s the AI tax.
The novel change: AI-assisted coding lowers the build barrier
Just as these models generate human language, they can generate and refactor code. What started as a novelty has exploded into one of the most valuable use cases. Developers get real productivity gains; the market is rougher for recent computer science graduates; and a meaningful share of many codebases are now AI-generated. In higher ed, this lowers the bar for building in-house, especially because most universities already have computer science faculty, IT staff, and students who can contribute to scoped projects. If Software as Service costs keep climbing, insourcing targeted workflows becomes viable again.
Will everything we build be great? No. Some projects will be brittle, insecure, or poorly maintained. Transitions are messy. But AI-assisted tooling makes it plausible to target narrow pain points—routing forms, lightweight advising workflows, content transformations, accessibility tools—where an internal solution can outperform a bloated enterprise product.
As Nik and I often note, sometimes AI highlights what was already broken. In edtech, it exposes a decade of vendor behaviors—lock-in, price ratchets, and stagnation—and offers a toolchain that makes alternatives credible again. The question for universities isn’t whether AI will show up in our software; it’s whether we pay the AI tax to vendors or apply the same technology to build what we actually need.