Adoption Outran the Org Chart
Instructure released research on July 21 surveying more than 1,100 higher education and K-12 educators, college students, and K-12 parents, and the topline number is stark: 90 percent of higher education students report using AI in class at least occasionally. That compares to 61 percent of higher education educators and 68 percent of K-12 educators who say the same about their own use. The gap between how often students reach for AI and how often the adults teaching them do is the real headline, and it is not a small gap. It is a structural mismatch between where a technology has already embedded itself and where the institution has built any capacity to manage it.
This is the pattern we have now seen across enterprise software categories repeatedly this year: usage arrives faster than governance, and by the time an organization writes a policy, the behavior it is trying to govern is already the default. Plenty of enterprise categories face the same lag, and higher education faces it with less budget flexibility than most, plus a workforce, faculty, that cannot be redeployed or retrained on the timelines a corporate L&D team can manage.
The Training Numbers Are the Story
Look past the adoption headline and the training data is where this gets uncomfortable for institutional leadership. 41 percent of higher education educators and 45 percent of K-12 educators told Instructure they received no formal AI training whatsoever. Only 11 percent of higher-ed educators and 8 percent of K-12 educators described their training as comprehensive. Confidence scores track accordingly: 2.60 out of 5 among higher-ed instructors and 2.53 out of 5 among K-12 teachers, both sitting below the scale's midpoint. These are not people who are resistant to AI. They are people who have been handed a tool their students already rely on without the institutional support to use it well.
Mary Styers, Instructure's Director of Evidence and Learning Strategy, framed the convergence across groups as the notable finding: "What stood out to me is how much agreement there is across these groups. Students, educators and parents see real value in AI, but they also want clear limits and better support." That agreement is the opportunity here. When students, faculty, and parents want the same thing, training and clear policy, an institution has a rare mandate to act instead of the usual three-way negotiation between constituencies pulling in different directions.
Everyone Agrees on the Same Worry
The survey's accuracy findings cut across every group surveyed. 65 percent of both higher education students and higher education educators, 60 percent of K-12 educators, and 69 percent of parents all flagged the same concern, that AI tools sound confident even when the answer is wrong. That is a rare point of consensus in a debate usually split along generational or role-based lines, and it suggests the anxiety about AI in classrooms is not really about whether to use it. It is about whether the people using it can tell good output from convincing-sounding bad output, which is precisely the judgment skill formal training is supposed to build and, per this survey, mostly is not building yet.
Despite the accuracy concerns, optimism ran high: 94 percent of higher education students and 88 percent of parents identified at least one reason for optimism about AI in education, and Instructure reports most educators did as well. The picture that emerges is not resistance to the technology. It is a population that has adopted a tool faster than its trustworthiness has been established, and faster than the institutions around it have built the guardrails to make that trust verifiable.
Where AI Is Welcome and Where It Is Not
Instructure's data also maps a consistent boundary across all three groups surveyed: AI is broadly accepted for support tasks, explaining concepts, brainstorming, and finding resources, and broadly rejected for grading or other academic decisions. That boundary matters for edtech vendors and IT leaders building or buying tools right now, because it tells you where the near-term addressable use case sits. Products pitched as replacing human judgment on assessment are going to meet resistance from the exact population that has otherwise embraced AI enthusiastically. Products that augment explanation, practice, and resource discovery are pushing on an open door.
Educators surveyed were specific about what they want to close the training gap: hands-on training and tutorials, clear institutional guidance and policies, and ongoing professional development, not one-off workshops. Chief Learning Officer Melissa Loble put the institutional obligation plainly: "AI use is already part of how students learn and educators work. The challenge now is making sure people have the training, judgment and clear expectations to use it well." For a company that sells the LMS this training would run through, that is also a fairly direct pitch for Instructure's own platform investments in AI governance tooling.
The Enterprise Read
Instructure has an obvious commercial interest in surfacing a training gap it can then sell tools to close, and that should temper how much weight any single institution puts on this specific data. But the underlying pattern, high organic adoption outrunning institutional readiness, is consistent with what we have seen reported independently across K-12 and higher ed all year, from teacher-adoption data in England to district AI policy fights in Florida. This is not a vendor manufacturing a problem. It is a vendor quantifying one that was already visible.
For technology leaders at companies that sell into education, whether LMS platforms, assessment tools, or corporate learning systems now competing for the same AI budget line, the actionable signal is the training gap, not the adoption number. Adoption is already won. The open market is in the professional development, policy templates, and confidence-building tooling that gets educators from a 2.5 out of 5 comfort score to something closer to fluent. Whoever builds that layer credibly, rather than just bolting another AI feature onto an existing product, has a real wedge into a buyer base that is currently underserved and openly asking for help.
