The teacher, the AI, and going off the road
Being a primary school teacher is 24 hours a week [...] and what people don't say is that it's six months a year, because between the holidays and the weekends...
That's what a former French president had to say about the job of a primary school teacher, at a November 2024 conference. This dream job, as we all know.
But try this little experiment at home:
- find someone who spouts the same kind of line
- then invite them to become a primary teacher, or a teacher of any kind
- watch what happens
Usually the conversation stops dead. "Uh... no, actually... uh." Strange, for such an enviable job, isn't it?
Ignorance, at first, might have excused our ex-president. But he keeps going, in the same ironic tone:
...I know, you have to prep your lessons... in kindergarten (laughter from the room)
In kindergarten, true enough, barely any papers to grade, not much "teaching" in the usual sense. From a distance, the prep looks light.
Except that's exactly where the job gets hardest. I know secondary school, I taught there. Primary isn't any simpler. Quite the opposite. At that age, what matters won't fit on a lesson plan: you're guiding the first steps of a whole new generation, where mistakes are hard to undo. Holding a room full of little kids together, building activities where each one moves at their own pace: that's the level that's least "plannable" on paper, and one of the most demanding.
So he knows. The irony borders on contempt. And no: he doesn't know.
He knows lessons have to be prepped. Obviously. But he doesn't know what that means, the same way ChatGPT doesn't understand what it generates.
This contempt has no party and no era. It's an old refrain that's been hanging around education forever. But it doesn't stop at the school gates. Today it's found a respectable new mask. AI.
The machine does the work, so we can make cuts.
It's one of those mantras you hear in the corporate world. Let's look at a case where a company took it a little too far:
Klarna, the big Swedish online-payments company, was bragging in early 2024: the equivalent of 700 customer-service jobs "replaced" by a full-AI chatbot. The truth was more mundane. The company was cutting costs, and AI made the perfect alibi. A year later, the CEO publicly backpedals: he admits quality had tanked, and starts rehiring.
The AI handled the volume but fell apart the moment something unexpected came up. Keep that detail in mind. We'll come back to it.
This same logic is now knocking on the door of schools and training centers.
Here, the alibi already has its coat of paint: the teacher becomes an "enhanced" teacher, thanks to the magic of AI. What that actually means: jobs you can cut, so classrooms packed even tighter, maybe a bit more remote learning too? The whole thing repainted as "innovation".
No.
I've built digital tools for teaching, built my own course platforms. That's exactly why I know what tech brings, and where it sometimes stalls. The time we save, we give back to the learners. We don't lighten the teacher's load. We equip them.
Back to the collapse of Klarna's AI. Why did an AI that used the most advanced tech of its day fall flat the moment things got unpredictable? It's not a question of power.
An AI learns from what has already happened. It knows the roads already mapped by heart, it can combine them, it can even improvise.
But the day it has to leave the road, there's no ground left under its wheels.
And no update fixes that. Neither does a beefier computer (yes, even a quantum one). The moment with no precedent, it can't see coming. It comes down to how computers fundamentally work.
A class, a training session, a room full of learners: you're always one step from leaving the road. Any teacher with a bit of mileage has lost count of the hours that drifted from the plan, however perfect it looked on paper.
Here's one I lived. I was running a Unix and Linux training for engineers at a government ministry. Right in the middle of a command-line demo, a question came from the group: "What's that keyboard shortcut you just did, to send your window to the other screen?"
The question hit home. I felt it, the group felt it. We dropped everything for half an improvised hour, swapping and testing our "top 10" keyboard shortcuts: the ones for daily work, and the ones to show off at geek parties ;)
Keyboard shortcuts, an AI spits them out in a second. What no AI could have done: sense that this question, at this exact moment, was worth dropping the plan for, and turn a side note into a shared moment for the whole group.
Those thirty minutes might seem trivial. They're not. One-way transmission is dancing swing alone while everyone else sits and watches. A good bit of improv is when we all step onto the floor together. Those precious moments carry real weight in a course.
If an AI couldn't hold up a simple support chatbot, who can seriously believe it'll hold a room full of learners mid-training?
Back to our ex-president.
He's got the same blind spot as that chatbot: he reads the map, "24 hours a week," but he's never driven off the road. He knows without knowing. He doesn't decide much anymore. But words like his still carry weight. They make contempt respectable, and they give cover to other people's cuts.
AI isn't the enemy. It's an excellent tool for the mapped-out, the high-volume, the repetitive. Let's use it for that, and only that. Not as the alibi for a budget no one will own up to.
The swelling bubble, layoffs repainted as progress: that's another story, and I'll get back to it. But for today, let's just keep this in mind:
What's really at stake lies off the road.
Where no machine can venture.