I Fear That The AI Adoption Gap In Schools Is Compounding

By Dan Fitzpatrick

Over Christmas, non-coders started building software in minutes. AI agents went from theory to product. Here's what that means for schools still on the fence.

The Questions Get HardThe surface-level read is that kids can build things with AI now. Fine. But the deeper issue is that our whole notion of what counts as a skill is starting to wobble. If AI can generate production-grade code from a plain English prompt, then the skill isn’t writing the code. The skill is knowing what to ask for, understanding the problem well enough to describe it precisely, and being able to evaluate whether what comes back actually solves it or just looks like it does. That is a completely different kind of literacy, and I'll be blunt, most curricula haven't started to wrestle with it. We're still drilling syntax when we should be drilling judgment. And this isn't just about computing. The same principle applies right across subjects.There’s a line I keep returning to when I talk with educators. Outsource your doing, not your thinking. An educator collaborating with AI to create a lesson plan on a Tuesday evening is outsourcing the doing. The professional judgment, the part where you decide what your class actually needs, how to sequence the learning, which misconceptions to head off, that stays with the teacher. A student pasting an essay question into ChatGPT and handing in whatever comes back is outsourcing the thinking. The entire purpose of the exercise disappears. Getting that distinction right is, I think, the central challenge for education right now. I was with a group of educators in Canada last week and we spent a good chunk of our session on exactly this question. Where’s the line? Do we even know yet?A Private Equity Firm’s FailureLet me share a story that puts some meat on this.Permanent Equity, a private equity firm, spent the whole of 2025 trying to build AI agents. Dozens of experiments, each one consuming over a hundred hours of work. Led by a PhD technologist they’d hired specifically for the job. In their 2025 annual letter, the firm’s leadership admitted that "the results have been shockingly poor compared to...