Generation is now free. Ask a model for a hundred logos, forty headlines, a landing page, a campaign, and they arrive in seconds, all competent, all plausible. The bottleneck was never supply. It was knowing which one is good.
That knowing has a name. Taste. And it is quietly becoming the most valuable thing a person can bring to a business.
Abundance Reprices Judgment
For most of my career, the cost was in the making. Producing the ad, writing the copy, building the page, shooting the video. Skill meant the ability to produce. So we hired and trained for production, and the people who could make things fast were the valuable ones.
That economy just inverted. When a model produces forty versions of anything on demand, the act of producing stops being scarce. What stays scarce is the eye that looks at forty options and knows which one to ship and why. The work moved from making to choosing.
Choosing well is harder than it sounds, because the machine gives you no help with it. A model can generate a thousand directions and rank them on engagement signals it scraped from the past. It cannot tell you which one is right for this brand, this moment, this audience, this bet. That judgment is a human standing in front of the options with a point of view.
Strip the romance out of the word and taste is just compressed experience. It is every campaign that worked, every one that flopped, every time the data said yes and the result said no, folded into an instinct that fires fast. You cannot prompt your way to it. You earn it by being wrong enough times to recognize right.
The Quiet Danger of Outsourcing the Decision
Here is the trap I watch teams walk into. The model produces something competent, and competent is seductive. It looks finished. It reads fine. So it ships, not because anyone decided it was the best version, but because nobody decided anything at all. The output became the decision.
That is the real risk of this era, and it is not the machines. It is people quietly handing over the one thing they were supposed to hold, the judgment about what is worth putting their name on. The model is happy to fill that vacuum with average, and average is exactly what it was trained to produce, because average is the center of everything it ever saw.
You can feel the result already. Content that is everywhere and forgettable. Pages that look the same. A flattening, where everyone uses the same tools to make the same competent thing, and nothing stands out because nothing was chosen. The cost of generation went to zero and so did the differentiation, for anyone who stopped exercising taste.
I have written that you should not automate the work that trains people, and this is the same problem one level up. If your juniors never struggle to make the thing, they never develop the eye to judge the thing. Automate the judgment too and you have a team that can ship infinite output and evaluate none of it.
How to Keep Taste in the Loop
Treat the model as a generator, never as the editor. It produces the raw material. A human decides. Make that division explicit on your team, because if you do not, the default is for the output to walk straight to publish unexamined.
Protect the work that builds judgment. The argument about which version is better, the critique, the willingness to throw away forty competent options to find the one that is actually right, that friction is not waste. It is where taste gets manufactured. A team that never says no to the machine is not efficient, it is asleep.
Give the editor real authority, not a rubber stamp. The person making the call needs the standing to reject competent work and the time to look hard enough to know it should be rejected. Rush that step and the model wins by default, because average is fast and judgment is slow. The companies that protect the slow part are the ones whose output will still feel like someone made it on purpose.
And develop a point of view on purpose. Taste is not neutral, it is opinionated, and opinion is precisely what averages cannot generate. The brands that will stand out in a world of infinite competent content are the ones run by people willing to say this, not that, and to be wrong in public until they are reliably right. I argued before that AI agents will not replace whole jobs, and this is why. The judgment at the center of the job is the part that does not automate.
The machine can make anything now. It still cannot tell you what is worth making. That decision is yours, and in a world where everything else got cheap, it is the most expensive thing you own.