AI Got Tenure Without a Job Talk

Picture a teacher in a staff meeting. Leadership has just announced an AI rollout. She raises her hand: "What problem is this solving for my students?"

In a room built around adoption targets, that question sounds like resistance. She gets signed up for a workshop. Or “By Friday, she’s on a performance improvement plan that includes a workshop on how to use AI. The workshop doesn’t answer her question.”

This week the European Commission published two reports on AI and digital technology in schools: Friend or foe? Evidence from the use of generative artificial intelligence in learning and teaching and the Eurydice report Digital education at school in Europe 2026. They're careful documents that take the risks seriously. The Commission's own summary says AI can improve performance without necessarily improving learning and warns that overreliance could affect how students develop critical thinking.

The reports also list teachers' lack of preparation and "skepticism" among the things hindering AI use in education.

That framing makes sense only if more AI use is already the goal. Somewhere along the way, AI got tenure without a job talk. No demo. No evidence. Apparently, being disruptive was enough.

THE BURDEN OF PROOF MOVED

We usually expect a new reading program to show evidence before it reaches every classroom. AI arrived with a different starting assumption: it's here, so how do we integrate it? That question quietly flips the burden of proof. The educator now has to justify holding back, and the tool gets waved through. Even the first report's title, Friend or foe?, casts AI as an actor with intentions, which makes its place in the room feel like a given.

Schools clearly have to respond. Four in ten young people in the EU aged 16 to 24 used generative AI for formal education in 2025. Students need to understand it, and schools need safeguards. Adopting AI as an instructional technology is a separate decision, and it should stay an evidence-based one.

So, reverse the order. Start with the learner and what education is meant to develop in them. Name the problem or opportunity. Then ask whether AI belongs, and under what conditions it serves the learning.

CURRICULAR OFFLOADING

In our last piece, we drew a line between cognitive offloading and cognitive surrender. The EU reports surface the same risk at system scale.

The reports suggest writing curricula may need to change because AI can now produce and refine sophisticated text. Follow that logic and you end up reasoning backward, from what a machine can output to what a human supposedly no longer needs to learn.

Writing produces text. The process of writing also builds reasoning, organization, synthesis and the discipline of revision. A tool that generates the final paragraph skips all of it.

Efficiency is useful when inefficiency is the problem. In learning, it often isn’t. Struggle, iteration, and revision can look like friction from the outside while being the exact opposite cognitive work the lesson was designed to produce. Before celebrating that AI made something faster, ask whether faster was actually the assignment.

Call it curricular offloading: an education system trimming its learning objectives because a machine can now perform the visible task attached to them. Scale makes it dangerous. A curriculum change reaches every student who passes through it, often for years. Before changing a curriculum, map the cognitive work the current learning experience performs, and decide which of those capabilities still matter whatever a machine can replicate. The measure stays human. What does the learner know, and what can they still do on their own?

SAME TOOL, DIVERGING RETURNS

The reports acknowledge that universal access to AI may create new disparities because students use it differently. That line deserves far more weight than it’s being given.

A student with strong foundations and sharp judgment can use AI to test ideas and pressure-check their own thinking. Hand the same tool to a student still building those foundations, and it becomes a shortcut around the hard part. The assignment still gets turned in. The grades may even be the same.

And the gap just got wider.

Equal access can produce diverging returns. AI can accelerate the students already ahead and deepen dependence among the rest. The Eurydice findings already show that students from disadvantaged backgrounds may have fewer opportunities to develop digital skills. Layer diverging AI returns on top of that and the gap compounds.

Waiting until AI is everywhere to study these patterns means studying them after they've set.

Equity work needs a sharper question. Who is becoming more capable because of AI, and who is becoming more dependent on it?

LISTEN TO THE SKEPTICS

Teachers who ask "why" are exercising the professional judgment we expect of them. The Commission itself calls for large-scale, longitudinal research on the long-term impact on learning and skills. With that much still unknown, doubt from the people closest to students reads as due diligence.

There's a contradiction here. We ask teachers to cultivate healthy skepticism toward AI in their students. But when teachers direct that same skepticism toward the technology itself, it becomes a “barrier.” Apparently, critical thinking is an educational outcome right up until the teacher uses it on the latest school-wide initiative.

Investigate teacher skepticism before trying to overcome it. Some of it will point to a training need; only 29.7% of EU teachers surveyed in TALIS 2024 had taken part in AI-related professional development. Some will point to a policy gap, or to a risk researchers haven't spotted yet. Asking teachers is how you find out which.

THE LEARNER-FIRST TEST

Before AI goes into any classroom, run the proposal through these 4 questions.

1. What should the learner be able to do? Name the capability this lesson exists to build.
2. Which educational problem are we solving? Describe it in plain language before any tool enters the conversation.
3. What does AI add, and what does it displace? Map the work the student would stop doing. Was that work the point?
4. Who benefits, and how will you know? Track access, differential benefit, patterns of use and changes in capability over time.

A proposal presented for consideration that answers all four has earned a pilot. Anything else isn’t ready for the classroom, let alone tenure. Anything else is paper weight.

INVITE AI IN LAST

The EU can stay ambitious about AI. Aim that ambition at the people education exists to develop and give real weight to the teachers standing closest to them. Then invite AI into the room and ask what, if anything, it can contribute. It can have a seat; it just doesn’t get the chair.

That sequence is the backbone of Fusion Compass, which starts with one question: are people prepared? If your institution is writing an AI strategy this year, start there. Reach out before your next rollout.