Ask educators what worries them about AI in the classroom and you'll hear about fit long before you hear about capability. The tools are powerful, and they can do real good in a classroom. The better question is whether they were built for learning or just pointed at it.
That's the question I'll dig into at CanvasCon next month in London, and it's one I get a lot. Educators need AI purpose-built for learning outcomes, not AI borrowed from somewhere else and thrown into a classroom. So what does that mean in practice?
Start where educators already work
We were clear from day one about what IgniteAI is. It's our integrated approach to AI inside Canvas: a coordinated set of AI-powered features and trusted partner integrations that show up exactly where an educator already is, in the gradebook, the inbox, or the quiz. It doesn't live in a separate window that has no idea what course it's in.
That's a harder standard than most generic AI tools hold themselves to, and educators tell us it's the right one. In a global survey of education leaders we ran, 72.2% named fairness, ethics, and institutional risk as their top concern about AI, ahead of raw capability. Nobody is questioning what the technology can do. They want to know it's being applied well and responsibly.
Use is ahead of structure
The adoption numbers show why that standard can't wait. The Digital Education Council's AI in Higher Education Global Survey 2026 drew more than 45,000 responses across 35 countries. It found that 88% of students and 77% of faculty now use AI in their work, both up 16 percentage points in a single year. Our own research, published this summer from a survey of 1,125 educators, higher education students, and K-12 parents, shows the same pattern closer to home: 90% of higher education students and 61% of higher education educators use AI at least occasionally.
Use is high. Structure hasn't caught up. The Digital Education Council found that just 29% of students believe their instructors can meaningfully guide them on AI, and only 31% of faculty feel involved in shaping institutional AI policy. In our research, 41% of higher education educators had received no formal AI training, only 11% called their training comprehensive, and the average educator rated their own confidence at 2.60 out of 5.
Melissa Loble, our chief learning officer, said it plainly when we published that research. AI use is already part of how students learn and educators work. The job now is making sure people have the training, judgment, and clear expectations to use it well.
The students feel this gap too. The Digital Education Council survey found 57% say they get inadequate direction on when and how AI belongs in their work, and only 28% feel their assessments measure the skills they'll need in an AI-enabled workplace. Nearly a quarter say they aren't permitted to use AI in any assessment at all, even as they use it everywhere else.
One product, deployed once, is the wrong model
Too many institutions still treat AI as a single product, deployed once, the same way everywhere. A tool trained on the open internet and handed to a student with no connection to the course's learning outcomes will behave exactly as generically as it was built.
EDUCAUSE's 2027 Top 10 IT Issues list puts this in sharp relief. It was compiled from 882 respondents, the largest voting pool in the survey's history, and AI displaced cybersecurity as the top issue in higher education IT for the first time ever. But the top-ranked concern wasn't "deploy more AI." It was "determine where AI adds real value." Mark McCormack, EDUCAUSE's senior director of research, said it this way: "This doesn't signal consensus as much as it signals urgency."
Purpose-built is measurable
This is where the distinction stops being rhetorical. Our 2026 Evidence Report, built with the nonprofit InnovateEDU and unveiled at SXSW EDU, analyzed 150 widely used classroom technologies. It found that 40% of purpose-built edtech tools have identifiable evidence of impact aligned to ESSA's evidence standards. Among the generic consumer technologies used in those same classrooms, the figure was 2%.
The appetite to get this right varies by region, and Europe is leading. The Digital Education Council found faculty intent to use AI at 89% across Europe, the Middle East, and Africa, compared with 67% in the United States and Canada, where the figure fell nine percentage points year over year. EMEA institutions aren't hesitant. They're asking the right question early.
What I want to talk about in London
Here's the conversation I want to have at CanvasCon Europe this year: was the AI that students are handed actually delivering the outcomes we're asking them to reach? Purpose-built is a design decision you can test. A tool that understands what a course is trying to teach will serve learners. One that doesn't know—or care—what course it's in never will.
