AI didn’t kill the work. It moved it somewhere harder.
The bottleneck used to be the hands. The hands now work for free. The bottleneck is the head.
Six months ago a junior could mock up forty options before lunch. Now an LLM does it before they finish coffee. The hands part of the job — making the artifact, rendering the comp, finishing the layout — has collapsed into seconds. That isn’t going to reverse. Production cost has rounded to zero, and it’ll keep falling.
What looked like the work was never the work. It only looked like it because production took so long.
The bottleneck used to be the hands. The hands now work for free. The bottleneck is the head — which option is right, which to kill, what the brand actually needs.
That’s a harder job than rendering. Rendering is mechanical. Judgment is not. Knowing which of forty options is the one, and being able to say so without flinching, takes a different muscle. It’s pattern recognition built over years of watching things ship and watching them work. Not transferable to a model. Not retrievable from training data. It has to be earned in a room with a brief and a deadline.
The first wave of AI was about producing more, faster, cheaper. The second wave will be about producing less. Fewer options. Stronger ones. The right ones.
The skill is no longer making fifty things. It’s killing forty-nine of them by Tuesday and standing behind the one.
What wins now is taste — which has been the quiet thing being measured all along, just disguised as craft. Craft was the visible part. Taste was the invisible part. AI exposed the difference. A team that can produce anything has to know which thing to produce. A team that can render any frame has to know the right frame. A team that can ship any campaign has to know the campaign that holds up.
The hands work for free now. Pay attention to whose head is in the room.