When production gets cheap, the thing that stays expensive is judgment.
That’s the uncomfortable truth sitting underneath a lot of the AI-and-creativity conversation. If anyone can generate a decent logo, a passable script, a solid-enough app interface in minutes, then “we can make things” stops being a differentiator. Every agency, every in-house team, every solo founder with a laptop can make things now. The gap that’s left isn’t skill in the old sense — it’s taste.
Taste is knowing which of the twenty generated options is actually good, and being able to say why in a sentence. It’s noticing that a headline is technically correct but emotionally flat. It’s the instinct to kill your best-performing idea because it’s right for the brief but wrong for the brand. None of that shows up in a training dataset as a clean pattern to copy. It’s built from years of making things, watching them land or flop, and updating your internal compass accordingly.
This is good news for people who’ve spent real time developing craft, and uncomfortable news for anyone hoping tools alone would get them there. AI can absolutely help someone develop taste faster — by exposing them to more references, more variations, more edge cases than they’d encounter on their own. But it can’t hand taste over pre-made. It’s earned, not generated.
For agencies, this reframes what clients are actually paying for. Not the asset. Not even the speed. The judgment that says: out of everything that could exist, here’s what should.
That’s not a skill AI is replacing. It’s the skill that decides what AI is even for.