Lately, AI has started to resemble a duck on water. On the surface, it looks smooth, elegant, effortless. The demos glide. The keynotes glide. The C-suite talks about four-hour workweeks and studying philosophy and picking up Italian in the time freed up by automation. Above the waterline, everything looks serene. Below it, the legs are paddling furiously, and a lot of humans are the legs.
This is the thing the serene version leaves out. The smoothness on top is produced by frantic motion underneath, and the motion is mostly people. People cleaning the data. People checking the outputs. People catching the errors before they reach a customer. People rewriting the prompt for the fifth time. People quietly fixing what the system got wrong so that, from the outside, it looks like the system did it right the first time.
I am not saying the duck is fake. The duck is real, and it genuinely moves across the water, and that is remarkable. The productivity gains are real. The capability is real. What I am saying is that the effortlessness is an illusion of surface, and the effort did not disappear, it moved. It moved below the waterline, where it is harder to see and easier to stop counting, which is exactly what tends to happen to work that becomes invisible.
And invisible work has a way of being undervalued, then underpaid, then unacknowledged entirely. The story we are telling about AI right now is a story about the surface, about the glide, about the freed-up time and the elegant output. The people doing the paddling do not appear in that story much, and when they do, it is usually as a transitional detail, the messy human step we will automate away any minute now.
But the paddling is not going away as fast as the story suggests. The systems still need enormous amounts of human judgment to work, the kind of judgment that does not show up in a demo because the demo is designed to hide it. Every smooth output you see has a substrate of human labor underneath it, and the smoother the output, the more tempting it is to forget the substrate is there.
This is my worry, and it is not a worry about the technology. It is a worry about the accounting. When we describe AI as effortless, we are not just being imprecise, we are erasing the specific people whose effort makes it look that way. We are building a story in which the machine does the work and the human is a nostalgic footnote, at exactly the moment the human is doing the load-bearing work of making the machine usable.
And the people paddling hardest are often the ones with the least standing to be recognized for it. The senior people get to talk about the strategy and the vision. The people below the waterline are doing the operational reality, the cleanup, the verification, the endless small corrections, and that work is being folded into the myth of the effortless machine rather than credited as the thing keeping the machine afloat.
So the reframe I want is simple. When you see the duck gliding, remember the legs. Not to diminish the glide, the glide is real and worth having, but to keep honest about what produces it. The effortlessness is a collective achievement that includes a lot of invisible human effort, and a story that keeps the effort visible is a story more likely to value and pay the people supplying it.
The duck glides because the legs never stop. The least we can do is remember the legs are there, and that they belong to people, and that the serenity on the surface is something they are producing, not something the water is doing on its own.