We built machines to do the things we found tedious. Calculations. Memorization. Routine writing. Pattern matching. The idea was simple: offload the mechanical to the machine, and free the human for everything else, creativity, judgment, meaning, connection. But somewhere along the way, the deal got reversed. We didn’t free ourselves from the machinery. We started competing with it.
The productivity reflex
Look at how we talk about AI today. The vocabulary is almost entirely industrial. Output. Throughput. Productivity gains. Time to value. Velocity. These are factory words. They are the words you use when you are measuring how fast a conveyor belt runs.
And increasingly, these are the words we use to measure ourselves. How many tasks did you ship this week? How many emails did you answer? How many words did you write? How much “deep work” did you put in, a phrase that, notably, treats human thought as a unit of production.
The reflex is everywhere. In schools, we measure students by activities completed, minutes engaged, modules finished. In knowledge work, we measure ourselves by output volume. In creative fields, we measure by frequency of posts. The metric of being human has quietly become the metric of being a productive machine.
What gets lost when efficiency wins
Here’s the strange thing: the qualities we used to associate with being deeply human, curiosity, reflection, struggle, slowness, the long pause before an idea forms, are the exact qualities that look like inefficiency on a dashboard.
A student who spends an hour staring out the window thinking about a problem before writing anything down is, by the standard EdTech metric, a low-engagement user. A writer who deletes a paragraph three times is, by the standard productivity metric, slow. A researcher who reads widely for months before knowing what their question is, is, by the standard funder metric, unproductive.
None of these people are doing nothing. They are doing the things that machines can’t yet do, and probably never will. But because our measurement systems can’t see those activities, we have started to treat them as waste.
We are now training AI on enormous quantities of human output, and then training ourselves to produce in ways that look more like that output. AI writes in clean, fluent paragraphs. So we start writing in clean, fluent paragraphs. AI produces confident summaries. So we start producing confident summaries. AI generates fast. So we start generating fast.
The mirror reflects, and then we adjust ourselves to match the reflection.
The risk isn’t that machines will become more like us. The risk is that, by spending so much of our time optimizing for the things machines do well, we will become less like ourselves. Less patient with ambiguity. Less comfortable with not knowing. Less willing to sit with a problem for the time it takes to actually think about it.
What’s worth protecting
There are things AI cannot do, not because the models aren’t good enough yet, but because the things in question are fundamentally not what machines are for. The slow accumulation of lived experience. The friction of caring about something. The act of changing your mind because something genuinely moved you. The willingness to be wrong in public. The patience required to sit with a learner who needs more time. The courage to ask a question you don’t know the answer to.
These are not productivity bottlenecks. They are not inefficiencies to be optimized away. They are the actual texture of being human, and they are worth defending, especially in education, where what we are really teaching is not how to produce more, but how to become more.
A different question
The conversation about AI usually starts with “What can it do?” A better starting question is: “What should we still do, even though it can?”
For educators, the answer might be: ask better questions. Hold space for confusion. Notice the learner who hasn’t said anything for a while. Have the kind of conversation that no machine can have, because it’s grounded in a specific relationship, in a specific moment, with a specific child.
For all of us, the answer is probably similar: protect the slow things. Defend the inefficient ones. Keep doing what is meaningfully human, especially when there’s a faster machine alternative.
Because the goal was never to compete with the machines. It was to be more human than they could ever be.