The Tool Arrived First

On June 24, Microsoft released the third edition of its annual AI in Education Report.¹ The headline number is that ninety-two percent of students and education leaders, and eighty-eight percent of educators, have already used AI for school work.¹ The survey was conducted by PSB Insights across six countries during the late winter, drawing on 3,345 K–12 and higher education respondents in the United States, the United Kingdom, Australia, Brazil, Japan, and Saudi Arabia.² It is one of the larger snapshots the field has of who is actually using the tool inside the building.

Read past the headline and the numbers form a quieter shape. Seventy-seven percent of students and fifty-three percent of educators say they have received no formal AI training.¹ Sixty-six percent of educators and fifty-two percent of students want their institution to provide that training at least quarterly.¹ Academic integrity, the report notes, is a leading worry for forty-one percent of students and forty-two percent of educators.¹

The pattern, once you stand back from it, is plain. The tool arrived first. The training is the thing everyone is still asking for, more than two years after the tool itself walked in.

A Census, Not a Forecast

It is tempting to read the report as a forecast. It is closer to a census. Microsoft is not asking whether schools will adopt AI. It is reporting that they already have, by the simple fact of who is in the building. Ninety-two percent of education leaders, the most institutionally cautious group in the sample, have personally used AI for school work.¹ Whatever the official policy on a given campus says, the tool is already in the staff lounge and the homeroom both. The next year of education policy is not about whether to admit AI. It is about how to behave once it is in.

This is a different problem than the one schools have been preparing for. A school that says no to a tool can build a year around the no. A school that says yes can write a syllabus. A school in which everyone is already using the tool, half quietly and without instruction, is doing something else. It is improvising in the dark. The pause in the third grade reading block where a teacher asks the model for a quicker explanation. The high school senior who asks for an outline before bed. The principal drafting a tense email to a parent at ten at night. None of those minutes are in the lesson plan. None of them are in the policy. None of them, in most buildings, are in the record at all.

What Training Cannot Reach Alone

The instinct of the report, and of nearly every district responding to it, is to fill the gap with training. Monthly workshops on prompt design. Quarterly refreshers on academic integrity. A literacy credential, like the one Microsoft just announced alongside the report with ISTE and ASCD, that says a teacher has learned how to use the model well.³ These will help. A teacher who knows how to scaffold a chat with a student before she writes is doing something different than a teacher who waves at the tool from a distance.

But training answers only the first half of the question. It tells the adult in the room how to think about the tool. It does not tell that adult what each child actually did with the tool on Tuesday morning. A monthly workshop does not show a teacher that her sharpest student paused over a counterargument for forty minutes before deleting it, or that her quietest student asked the model the same clarifying question six different ways. The workshop educates the teacher in the abstract. The Tuesday morning, in most schools, is still a sealed envelope.

A future built on training without a record will keep relitigating the same debate the report has just measured. Are students learning more, or less, with AI in the room? Without a way to look inside the minutes, every district will answer from its own anecdotes. The school that wants its training to compound will need a second thing alongside it. A record of how the work was actually done. The draft saved next to the final. The pause in the writing surface logged next to the score. The conversation with the model captured and made available to the teacher, not as evidence of cheating, but as the ordinary texture of how a child learned this week.

The Year After the Census

What Microsoft published on June 24 is the closest thing the field has to a snapshot of where it actually stands. Ninety-two percent in. Most still unsure of how. Almost everyone asking for help. The next move is not a vote on whether AI belongs in the classroom. That vote has already been counted, quietly, by the people inside the classroom. The next move is whether the building can begin to see what the people inside it are doing with the tool, before another year of training is poured into a year no one has any way of reading back.

A teacher who knows how to use AI is a real gain. A teacher who can see how each of her students used AI this week is a different kind of gain. The first is a credential. The second is the start of a profession that knows what it is teaching into.

If the survey found that ninety-two percent of your school is already using AI, what would the building actually need to learn from that fact before September?