A marketing team asks for launch headlines and gets two hundred by lunchtime. Some are clumsy, most are competent, a handful are good, and one is right. The team spends the afternoon arguing about which one that is. Nobody notices that the argument, not the writing, has become the actual work.
This is what happens when production gets cheap. For most of the history of office work, the scarce thing was output: someone had to sit down and make the deck, the draft or the code. Seniority tracked that scarcity. The senior person could make more of it, faster, with fewer mistakes. Now an agent can produce competent first versions of almost anything in minutes, and the scarce thing has moved. What matters is the ability to look at a pile of plausible options, say which one is good and why, and say no to the rest.
Call it taste. It's an awkward word in a business setting, because it sounds like an opinion about wine. It is also what now separates the people who make work better from the people who make more of it.
Cheap output creates more choosing
The Jevons paradox, first observed with coal, says that making a resource cheaper to use often increases how much of it gets used. Something similar is happening with drafts. When a proposal takes a day to write, a sales team sends a few careful ones. When it takes ten minutes, they send dozens, each in five variations, and someone has to decide which variation goes out.
The bottleneck doesn't disappear. It moves to selection, and selection was never anyone's formal job. It happened silently, inside the head of the person doing the making, who discarded nine ideas before writing down the tenth. Now the nine discarded ideas arrive fully formed and nicely formatted, and somebody has to discard them on purpose.
That somebody needs to know what good looks like for this customer and this moment. Producing an option no longer takes any special knowledge. Picking the right one takes all of it.
What taste looks like at work
Strip away the mystique and taste in a work context comes down to a few concrete abilities. Knowing the audience well enough to predict how they'll read something. Knowing the history: which approach was tried two years ago and why it failed. Knowing the constraints nobody wrote down, such as the client who hates the word "solution" or the regulator who reads every footnote. And holding the goal clearly enough to tell a clever option from a useful one.
Consider an engineering lead looking at three working implementations of the same feature, all generated in an hour. All three pass the tests. One adds a dependency the team has spent months trying to remove. One is elegant and will baffle whoever maintains it next year. One is plain, slightly longer, and fits the way the rest of the system is built. Choosing the third is taste. It doesn't show up in any test, and it's worth more than the code.
Or a recruiter reading ten AI-drafted job adverts. Every one is grammatical and on-brand. Only one sounds like a place the right candidate would actually want to work, because it mentions the thing that makes the job hard instead of hiding it. Spotting that takes someone who has hired for the role before and remembers who stayed.
Saying no is the hard part
Recognising the good option is half the skill. The other half is rejecting the rest, out loud, in a room full of people who liked them.
That is harder than it sounds, because abundance changes the politics of a meeting. When each person used to bring one idea they had worked on for a week, killing it was painful and rare. Now everyone arrives with twenty, generated between other tasks, and each has someone attached to it. The cost of producing an idea has collapsed. The social cost of rejecting one hasn't moved at all.
Teams without someone willing to say no drift towards the worst outcome available: shipping everything. The website rotates four headline variants. The product gets every feature the agent could build. The quarterly report includes every chart because each one was free to make. Nothing is exactly wrong, and the whole is worse than a smaller, edited version would have been.
Saying no well is a skill of its own. A bare "no" teaches the team nothing and leaves them generating the same near-misses next week. A "no, because our buyers are finance people and this reads like it was written for designers" turns a rejection into a rule others can apply before the next batch arrives. Good editors explain their cuts. Over time, the explanations travel down the team, and fewer weak options reach the table in the first place.
It also helps to decide what you are choosing for before the options appear. A team that agrees in advance that the launch headline has to work for existing customers as well as prospects can discard half the pile in minutes. A team that starts reading without a criterion ends up choosing by mood, or by whoever spoke last.
An editor is defined less by what they publish than by what they kill. Every organisation now needs more people who think like editors, and most have spent decades training their staff to think like producers.
Career ladders still reward volume
Most promotion criteria were written for a world where output was hard. They reward the person who shipped the most or closed the most. It makes less sense when a junior with a good agent can outproduce a senior without one, and when the senior's real contribution this quarter was the three things she stopped from going out.
Stopping things is hard to see on a performance review. Nobody writes up the campaign that didn't run or the feature that wasn't built. So the people with the best judgement can look, on paper, less productive than the people generating the most noise, and organisations risk promoting exactly the wrong instinct.
The fix starts with making selection visible. Ask people to explain why they chose one option over the others, and treat a good explanation as work product. Give the role of final judge to named people, instead of letting decisions emerge from whoever has the most stamina in meetings. And when someone kills a weak idea early, record it as a contribution.
Where taste comes from when the apprenticeship shrinks
There's an awkward problem underneath all this. Taste has always been built through making. The senior designer has good judgement because she produced hundreds of mediocre layouts and learned, slowly, why they were mediocre. The experienced lawyer spots a weak clause because she once drafted one and watched it fail in a dispute.
If juniors now start their careers selecting instead of making, they skip the stage where judgement forms. They can tell you which option the senior picked without knowing why. Give it a few years and the organisation has plenty of people who can generate options and too few who can choose between them.
The answer is to keep deliberate making in the early years, even when it's slower than asking the machine. Have juniors write their own first draft before they see the generated ones, then compare. Have them defend a choice in front of someone senior, and lose the argument now and then. It looks inefficient, and it remains the only reliable way anyone has found to grow judgement.
Meanwhile, look at your own team and ask a narrow question. When twenty options land on the table tomorrow, who is allowed to pick one and bin the other nineteen? If the answer is nobody, or everybody, that's the role to fill first.
