AI in education: what I see in the field

Everyone talks about AI in education. Conferences multiply, white papers pile up, vendors release “AI solutions” every week.

But how many people actually use it? Daily? With real learners, real instructors, real constraints?

What I’ve tested

For the past year, I’ve been progressively integrating generative AI tools into the educational processes I manage. Not in a lab — in the field, in schools, with teams that don’t always have a tech appetite.

Here’s what I’ve observed:

What works

• Course material generation: a significant time-saver for instructors, as long as you prompt well and always review
• Assessment support: not for auto-grading, but for generating rubrics, criteria, and feedback suggestions
• Research and synthesis: summarizing a 20-page article into 3 key points — that’s where AI truly shines

What doesn’t work (yet)

• Leaving learners alone with ChatGPT: without framing, it’s sophisticated copy-paste
• Promising it replaces the instructor: it replaces nothing. It augments. Big difference.
• Deploying without training: an unaccompanied tool is a dead tool

My take

AI in education is neither the revolution being sold nor the danger being waved around. It’s a tool. Powerful, imperfect, and requiring method.

The real challenge isn’t technological — it’s human. It’s change management. It’s convincing a 55-year-old instructor that this tool can help without replacing them.

And that, no AI can do. It’s fieldwork.