AI is now part of the edtech stack — and schools are repeating the mistake they made in every wave before

The latest school tech wave is running into the same wall as the last one. The device or software gets installed, the launch email goes out, the dashboard lights up, and then nobody can point to better work, better teaching, or better student outcomes. AI changes the excuse, though. This time the tools are genuinely strong, and the gap is still there.

That matters because the Department of Education has stopped treating all screen time as the same thing. Its new guidance separates recreational tech from instructional tech and says the old usage metric is lousy. More importantly, it pushes schools and vendors toward something harder: show learning outcomes, build evidence into renewals, and disclose what products can’t do. The message is simple. Use technology if you want, but prove it helps.

The source article argues that schools already lived through this mistake during the pandemic-era push for laptops, interactive whiteboards, and other software buys. The hardware showed up. The training and the teaching model often did not. Career-exploration tools offer a clean example. States invested in them, some with help from top game designers, but schools still had to carve out class time, hire coaches, and line up employers for internships or apprenticeships before the tools meant much.

AI tutors are already showing the same pattern in smaller form. In a two-year randomized trial across 18 Tennessee middle schools, 96% of students tried the tutor at least once, but they used it in only 17% of the moments when they had made a mistake. Students often skipped the help that was sitting right there. A companion experiment found better results when software forced students to slow down and review their work, with the strongest gains coming when AI was built into a mastery-based workflow. The lesson is blunt: a capable tool is not the same thing as a learning outcome.

And the implementation problem is getting bigger, not smaller. Alabama plans to require a computer science course including AI instruction for graduation starting with the class of 2032. Idaho has told its education department to build a statewide AI framework. FutureEd is tracking 77 AI-in-education bills across 27 states this session, while the OECD plans to test media and AI literacy in PISA in 2029. None of that works without adults who know how to teach the thing. That is the real bottleneck.