On 27 August, OpenAI announced the expansion of its offering for teachers to new US school districts. The interesting point is the targeting: the tool is aimed at teachers, not students. It is a strategic choice, and it partly sidesteps the debates that accompanied the arrival of AI in classrooms.
We have covered what works and what doesn't at school and the experience aimed at teenagers. Here is the third instalment, and it targets the other side of the desk.
Why targeting teachers is smarter
A tool aimed at students immediately triggers the difficult questions: cheating, cognitive atrophy, protection of minors. A tool aimed at teachers sidesteps most of these debates.
And it addresses a real, documented problem: the administrative burden. A considerable share of teachers' working time is absorbed by tasks that are not teaching — preparing materials, differentiating an exercise for three levels, writing report comments, producing documents for the administration.
These are exactly the tasks where these tools are most useful: textual, repetitive work with low factual stakes, where the teacher remains the judge of the result.
What it can actually bring
Differentiation. Adapting the same exercise to different levels is a time-consuming task that many teachers give up on for lack of time. Making it quick directly benefits the students who are struggling most.
Preparation. Generating variants of a problem, producing a detailed answer key, coming up with comprehension questions.
Time given back. This is the central argument, and it is an honest one. Every hour taken away from admin is an hour available for what cannot be delegated: being with students.
Who pays next? A rollout at a preferential rate creates dependency. A district that has reorganised its work around a tool is in a weak position at renewal time, and we saw this summer that prices are going back up.
What happens to the data? Student work, report comments, individual situations are sensitive data concerning minors.
Will the time saved go to students? A productivity gain can fund more teaching time, or justify larger classes. This is not a technical question; it is a budgetary one.
The least discussed risk
There is a consequence that is rarely mentioned, and it concerns the profession itself.
Preparing a lesson is not just producing materials. It is the moment when a teacher reorganises what they know, anticipates difficulties, decides what matters. This work transforms the person doing it, as we wrote in our article on time saved.
Would a beginning teacher who had never built a lesson sequence from scratch learn their craft the same way? The question is not rhetorical, and it also applies to learning to code: when the tool produces the result, the path that formed you disappears.
What to take away
This is probably the most defensible school rollout we have seen. It targets a real burden, it leaves the adult as the decision-maker, and it does not directly touch students.
What will decide its value is not played out in the tool but in how institutions use it. A teacher who gets five hours back per week and devotes them to their students represents a considerable gain. A system that notices this gain and concludes that five more students can be added per class cancels it out entirely.
Technology will not settle the matter between these two outcomes. The budget will.