You’ve been here before. The district pushes out an AI tutoring platform: personalized instruction, instant feedback, scalable to every kid in the building. You send the link, run a quick walkthrough, and wait to see students dive in. A few weeks later, you pull the usage data and most of them haven’t touched it since the first day.
New research confirms what teachers have quietly observed: access to an AI tutoring platform does not translate into use. Researchers tracking major AI platforms found that students were not using the tools as much as ed-tech companies expected, and were not asking the kinds of questions that actually lead to learning. The adoption gap is real and documented.
This is not a technology problem. It’s a design problem, one teachers can actually solve.
The Monday-morning move
The tools that get used share one feature: they’re embedded in a task students are already required to do, not offered as an optional resource alongside that task.
Here’s how to redesign one assignment this week. Setup time: about five minutes.
Pick something you’re already assigning. No new unit, no extra prep.
Choose one specific moment inside that assignment (before a problem set, after a first draft, midway through a lab) and make the AI check a required step in the sequence. Students move to the next section after they’ve completed it.
Give them a specific prompt, not a vague invitation. Not “use it if you’re stuck.” Something like: Enter your answer to question 3 and ask the tool what you got wrong. The more specific the prompt, the less students have to invent their own entry point.
At the end of the week, show the usage data to the class, not to catch anyone who didn’t engage, but to normalize what engaged use looks like. When students see a usage dashboard together, “I tried it” becomes a shared reference point rather than a private experiment.
Difficulty label: works in a 40-minute period as long as students have individual logins and a platform that runs on school devices without sign-in friction.
The honest caveats
This approach has real conditions:
The platform needs a usage dashboard you can read in two minutes. If you can’t see whether anyone clicked, you’re designing blind.
Students need to reach the tool without creating a new login in the middle of a task. Asking a student to switch accounts during a timed writing prompt is not a checkpoint; it’s a reason to give up.
And the tool needs to do something genuinely useful at the specific moment you’re inserting it. If the fit feels forced, the checkpoint becomes busywork.
There’s a class this won’t suit. Fragmented periods, shared devices with sign-in friction, or a platform that’s still unreliable on school networks: any of these can make a required checkpoint more frustrating than instructive. If that’s your situation, the principle is still right, but the environment has to be ready first.
If you only do one thing
Convert one optional AI step into a required one this week, with a specific prompt, and check the usage data by Friday.