I work in AI edtech and I've also stood on the other side of that conversation, in the classroom, staying after class with a learner who has failed the same class twice and was starting, slowly, to get it.
Which is probably why I notice something in almost every demo I've sat through or run myself: the questions being asked are almost never the ones that would actually tell them whether a tool is worth using.
They ask about features, integrations, pricing and implementation timelines. All reasonable. But none of those questions get at how a product was built or whether the people building it have thought seriously about how learning works.
Joy Delizo-Osborne at Student Achievement Partners recently put a name to something I've been feeling for a while: technopragmatic. Writing in The Hechinger Report, she argues that the AI-in-education conversation has been dominated by what the technology can do, with far less attention paid to what learners need to develop critical thinking or to the human relationships that make difficult learning possible.
She's right. And I'd add one thing: that imbalance isn't just a messaging problem. It shapes what gets built.
Here's something the demo won't tell you: most AI tools in education are built around product roadmaps, not learning science.
I've watched parts of the higher ed and language learning space run headfirst into this tension already. Duolingo's AI-first push is one example. Some of the innovation there is genuinely interesting. But the backlash revealed something important: learners and educators can tell the difference between technology that makes content faster to produce and technology designed to support learning. Those are not always the same thing.
A product can become more scalable, more automated and even more impressive in a demo while quietly becoming less thoughtful about the actual cognitive work of learning.
If I were sitting in the buying seat right now, there's one question I'd ask to actually live out what technopragmatic means in practice:
Can you show me where and how your product was built around educational research and learning science?
Not whether they believe in it. Show me where it lives in the product and how it shaped what you built.
A team that has done that work will answer without hesitating. They'll point to specific decisions: why feedback is structured a certain way, how the product handles a wrong answer, what research informed the pacing. A team that hasn't will tell you about their features.
The tools that support learning are built around how confusion works, how understanding develops over time and what it takes to help a learner move from getting the right answer to knowing why.
That work is slower and harder. It doesn't always make for a flashier demo. But it's the difference between a tool that enhances learning and one that simply simulates it convincingly.
Ask the question. The answer will tell you everything the slide deck won't.
Because at the end of this - past the demos, procurement decks and product roadmaps - there's a learner who has already failed this once and is trying again.
What they get should be built by people who understand what that actually takes.
That's worth asking about.