Every back-to-school email carries the same subtext: here are the new systems, the new compliance audits, and the new documentation requirements. None of them take time away. They’re simply additions. And when a principal or superintendent reads that sentence, they believe it. But teachers know what’s happening: the new thing doesn’t replace the old thing. It lands on top. The hours to grade don’t shrink because there’s a new platform. The emails don’t stop. The meetings don’t cancel.
The result is what everyone sees: teachers at 6 PM still working. Teachers who haven’t eaten lunch. Teachers grading on Sunday. The assumption is that the work itself increased. And it has. But that’s not the whole story.
In July, a team at Stanford published research that points at something deeper. They ran a randomized trial with Turkish middle and high school teachers. Some got an AI teaching assistant. Others didn’t. The AI did exactly what it promised: it generated lesson materials, drafted exercises, created syllabuses. Teachers used it. Then the researchers measured what happened to students.
Student motivation dropped. Not by a tiny margin: by 0.11 standard deviations, enough to matter. Students rated the AI-assisted courses as less enjoyable, less interesting, less important. And for students whose teachers were already struggling, it got worse. Their test scores fell. Their confidence fell.
The researchers framed this as a principal-agent problem. The principal (school leadership) wants better outcomes, so it gives the agent (the teacher) a tool to be more productive. But what the principal doesn’t account for is that the teacher’s actual job (being present, being curious about what a student doesn’t understand, being the reason a student shows up) doesn’t get more productive just because there’s less paperwork. It gets *less* present.
The study also found that teachers who used AI heavily at baseline became more pessimistic about its impact. The ones who’d used it less became more optimistic. The researchers wondered if there’s an arc: you discover how powerful AI is, you’re excited, and then you live with the reality: the tool’s job and your job aren’t the same job. You generate materials more efficiently. Your students learn less, at least partly because you’re not as much *with* them.
This isn’t a critique of individual teachers. It’s a structural problem, and it has a structural cause: **you’ve given teachers a tool and taken away their time.**
When teachers can’t find a sixty-minute block to sit with a student who’s lost, they use AI to draft an explanation instead. When there’s no space in the calendar for teachers to talk to each other about what’s difficult in their classroom, they grab a template instead of building something together. When every hour is accounted for (teaching, grading, mandatory meetings, compliance documentation, new platform training), creativity dies. And when creativity dies, automation becomes inevitable.
Consider this. A teacher with two hours of unaccounted time a week can notice that half the class is checking out during discussion, can sit down and think about why, can redesign that section. A teacher with zero hours can’t. A teacher with protected planning time can sit with a colleague and say, “Here’s what I’m trying. Am I missing something?” A teacher with no protected time reaches for AI. The constraint isn’t the tool. It’s time.
The UK published research this summer showing that 68% of university students think AI skills are essential to their future. Only 48% feel like anyone is teaching them those skills. The gap isn’t knowledge. It’s attention. Someone showing up, taking an hour, thinking hard about what matters and why. That can’t be automated.
Schools don’t say this out loud, but if you keep compressing teacher time until there’s no space for thinking, creativity, relationship, or judgment, you’ve built a system that wants to be automated. You’ve made your own teachers obsolete by making it impossible for them to do the thing only humans can do.
For your leadership team, this means every new initiative you add has a cost that doesn’t show up in a budget line. It costs teacher time. And you’ve already spent that budget. You’ve spent it on compliance, on grading platforms, on data dashboards, on documentation. Whether AI belongs in your schools isn’t the issue. What matters is whether you give back the time that has to come from somewhere, or let your teachers disappear into automation.
If you want teachers to be teachers (to think, to create, to respond to what students need), you have to give them time. Not tools. Time. Because right now, your email is a tool. Your platform is a tool. Your documentation system is a tool. Your meetings are tools. And they’re all taking the only real resource teachers have left: hours in the week.
Start here: audit everything on a teacher’s calendar and everything in their email inbox. Ask: does this have to happen? Does it have to happen this week? Does it have to happen this way? Every answer that’s “yes” belongs there. Every answer that’s “maybe” or “because we always do it this way” is a line item you can cut and give back to teachers.
Because if you don’t, you’re not just unhappy teachers. You’re teachers who’ve accepted that their job is to manage systems, not to teach. And at that point, a system can do it.
Takeaway for Leaders: Before you add a new initiative, ask: what teacher time does this require, and what am I taking away to make room for it? If you can’t answer the second part, don’t do the first.
If you’re thinking through structural barriers to teaching, my book goes deeper. The AI Doesn’t Know Your Students is available on Amazon and at shouldiuse.ai/book.
David Jacobson is a high school history teacher. He writes about AI, education, and the messy intersection of the two at shouldiuse.ai.
