The Number Under The Number

On Tuesday, at InstructureCon 2026 in Louisville, Kentucky, Instructure released the results of a survey it fielded last month of one thousand one hundred and twenty-five students, teachers, and parents across K-12 and higher education.¹ The headlines moved to the largest number in the deck. Ninety percent of higher-ed students report using AI in class at least occasionally.² Sixty-eight percent of K-12 educators do.¹ Seventy-three percent of K-12 parents say their child does.¹

Any one of those numbers, three summers ago, would have been the story of the year. This week, they moved through the news cycle almost silently. The word adoption has finished the arc of surprise.

Under the adoption numbers is a smaller one that has surprised almost no one, and that ought to have. Forty-five percent of the K-12 educators surveyed said they had received no formal AI training.³ Forty-one percent of higher-ed instructors said the same.² All three of the shared concerns the survey surfaced, overreliance, loss of critical thinking, and academic integrity, are downstream of that number.⁴ They are the specific worries of adults asked to guide a technology they have not been formally taught to see.

That is the number to sit with going into the fall.

What Forty-Five Percent Really Says

Forty-five percent is a number about time. The tools moved fast. Districts moved fast to buy them. Professional development did not. The teacher who will guide her students through their first AI-mediated writing assignment in September did not, statistically speaking, spend her July at a workshop about it. She spent it thinking about her classroom, her kids, her curriculum, and the seventy other logistical questions the summer always demands.

She will meet her students already knowing, quietly, that many of them use these tools more fluently than she does. She will be asked in the same year to teach with them and about them. She will not have had the professional grounding the last few big changes in the classroom (calculators, the internet, one-to-one devices) received before they became ubiquitous.

What The Gap Actually Looks Like

Read the survey more carefully and the gap is not really between users and non-users. Almost everyone uses. The gap is between the people who press the button and the people who can name what the button is doing to their thinking, to their students' thinking, to the classroom.

Ninety percent of students use AI. A comparable share of the adults do, too. What separates the two groups is not the tool. It is the vocabulary. On that axis, the survey is quieter but clear. Sixty-five percent of college educators and college students said they worry that AI sounds confident when it is wrong.⁴ The other named concerns cluster in the same place. Overreliance. Loss of critical thinking. Academic integrity. All of them live in the adult mind first. The student, statistically, is not going home worrying about any of them. That is not a failure on their part. It is the reason there is a subject called school in the first place. The adult's job is to name what the student is doing before the student is old enough to.

You cannot name what you cannot see.

What The Fall Requires

The Instructure survey is a good survey. It is also, in the way industry surveys often are, a description of the problem that stops one step short of the shape of the solution. The last sentence of the accompanying press release calls for institutions to build the insight, support, and guardrails to use AI in responsible, human-centered ways.⁴ The word to underline in that sentence is insight.

Insight is not another training module. Insight is not another summer all-hands. Insight is the specific, granular ability of a teacher to look at a student and see what happened between the prompt and the paper. The pause the student took. The revision she made. The sentence she borrowed and did not put in her own words. The moment she gave up and asked the model to finish. Insight lives in the trace of the work, not the finished work itself.

The forty-five percent will close, of course. Districts will run trainings. New teachers will graduate having grown up with these tools. But it will close slowly, over years, in a moment that will not wait years. What the teacher can be given in the meantime, before the training arrives, is a view of the work she was otherwise going to have to guess about.

Where Koan Sits In This

This is what we mean when we say Koan is instrumentation. Aidan, our AI tutor, does help produce and scaffold. What matters more, in the current moment, is the record. Every revision a student made. Every pause longer than thirty seconds. Every sentence that arrived on the fourth try after three false starts.

A teacher who has never been formally trained in AI can still, opening Koan on a Monday morning, see what her students actually did on Friday afternoon. That is what closes the gap in the short run. The vocabulary follows the seeing.

The Number To Watch

Ninety percent will be the number the press moves toward next year. It will climb to ninety-five, then to ninety-eight, and eventually it will disappear from the reporting because it will be assumed. Forty-five percent is the number to watch. It is the one that will decide whether a decade from now the phrase AI in schools describes a technology the adults understood in time, or a technology that was allowed to shape a generation while the people responsible for that generation caught up.

The distance between those two futures is not a technology question. It is a visibility question. And it will not resolve itself.

What would you need to be able to see in September, before you would trust yourself to guide a student through this?