Adoption Is No Longer the Question
Microsoft has published the third edition of its annual AI in Education Report, and the headline numbers confirm what most educators already sense: AI is everywhere in schools. According to the study, 92 percent of students and education leaders have used AI for school purposes, and 88 percent of educators have done the same. More than half of education leaders say their institutions are implementing or scaling AI, and use has risen sharply over the past year across students, educators, and leaders alike. Whatever debate once existed about whether AI belonged in education, the data says that debate is over in practice.
We read these figures as a decisive shift in the nature of the problem. For two years the conversation centered on encouraging or resisting adoption. That framing is now obsolete. When more than nine in ten participants already use the tools, the question is no longer whether AI enters the classroom but whether it does so with competence, guardrails, and purpose. As Matt Jubelirer, Microsoft's general manager of education marketing, put it, educators around the world are embracing AI as a classroom ally, and they are now asking not if, but how to make the most of it. The word how is where the hard work begins.
The Training Gap Is the Real Story
The most alarming finding is the gulf between use and preparation. While 87 percent of educators and leaders agree that AI competency matters for students' futures, 77 percent of students and 53 percent of educators report that they have received no formal AI training at all. People are using powerful tools daily with no structured instruction on how to use them well, judge their output, or understand their limits. Demand for that training is clear, with two-thirds of educators wanting AI instruction at least quarterly, but supply has plainly not kept pace with the speed of adoption.
This gap is the finding education leaders should lose sleep over. Unstructured use of AI without training is not a neutral state. It is how bad habits form, how misinformation propagates, and how the benefits of the technology accrue unevenly to those who happen to already know how to wield it. We would argue that the risk in education is no longer under-adoption. It is ungoverned adoption, where students and teachers improvise with capable systems and no shared standard of practice. Closing the training gap is not an enrichment activity. It is the prerequisite for AI in schools to do more good than harm.
Concerns Have Matured
The concerns captured in the report have grown more specific and more grounded than the abstract fears of a year ago. Academic integrity leads the list, cited by roughly 41 percent of students and 42 percent of educators, a striking convergence that suggests both sides of the classroom recognize the same tension. Students worry about integrity nearly as much as their teachers do, which complicates the lazy narrative that learners simply want to cut corners. The reality is that both groups are wrestling with what honest work looks like when a capable assistant is always within reach.
We find the maturation of these concerns encouraging, because vague anxiety is hard to act on while specific worries can be addressed with policy and design. Integrity concerns point toward clear assignment guidelines, assessment redesign, and transparency about when AI use is permitted. This is precisely the kind of governance that a training program should teach and that institutional policy should codify. The fact that students share the concern means the solution can be built with them rather than imposed on them, which tends to produce norms that actually hold in practice rather than rules that are quietly ignored.
New Tools Aimed at the Gap
Alongside the research, Microsoft announced a wave of features designed to move schools from experimentation toward structured use. For educators, Unit Plans in Teach offers AI-assisted, standards-aligned lesson planning, and new student AI guidelines in assignments let teachers define exactly how AI may be used on a given task. For students, Copilot Notebooks provides an AI study workspace at no additional cost, and a Study and Learn Agent offers research-based guidance with interactive practice rather than simply handing over answers. The through line is scaffolding, tools that structure how AI is used rather than leaving it open-ended.
The design intent we notice is a shift from generic chatbots toward purpose-built, pedagogy-aware tools. A general assistant will happily write a student's essay. A study agent built on learning science is meant to guide the student to write it themselves, which is a materially different product with a different effect on learning. Microsoft also introduced an Elevate for Educators program offering a free AI literacy credential co-created with the education organizations ISTE and ASCD and grounded in European Commission and OECD frameworks. Whether these tools deliver depends on implementation, but the direction, structure over open-endedness, is the right response to the training gap.
What Education Leaders Should Do Now
For superintendents, principals, and chief information officers in education, the report is less a set of statistics than a to-do list. Adoption is handled; the market has taken care of that. The unfinished work is competency and governance, and it falls to leadership to build both. That means investing in sustained teacher training rather than one-off workshops, writing clear and enforceable policies on acceptable AI use, and redesigning assessment so that integrity is protected by design rather than by prohibition. The institutions that treat these as urgent priorities will pull ahead of those still debating whether AI belongs in the classroom.
The broader lesson applies well beyond schools. Microsoft's data captures a pattern visible across every sector adopting AI, where capable tools spread far faster than the training and governance meant to accompany them. Education is simply an early and vivid case because its users are young and its stakes are formative. We would urge education leaders to treat the training gap as the defining challenge of this phase, because a generation is learning to use these tools right now, with or without guidance. The choice is not whether they use AI. It is whether they learn to use it well.
