Selection is not merely preference. It is disciplined rejection. The selector needs criteria, taste, context, and willingness to discard work that is competent but wrong for the purpose.
This becomes especially visible in creative work. If a system can generate one hundred musical ideas, images, taglines, or strategic options, abundance can become noise. The professional who can say 'not this, not this, this one, and here is why' creates value by compressing possibility into commitment.
Selection also carries responsibility. Once possibilities are cheap, 'the AI suggested it' becomes a weak explanation. The human who selects an output makes it part of the world. Selection is therefore both an aesthetic and an ethical act.
The economics of choosing
For most of history, generation was expensive. A company could not cheaply commission one hundred campaign concepts, one hundred product names, one hundred strategic scenarios, or one hundred musical arrangements. Generative AI changes the cost curve. Possibility becomes cheap enough to create a new problem: abundance overwhelms attention.
That makes selection an economic activity. The selector must know what fits the objective, what is distinctive, what is feasible, what violates constraints, what creates downstream risk, and what deserves scarce organizational attention. Selection is not preference alone. At its best, it is compressed expertise.
This explains why taste matters more, not less, in generative environments. Taste is the ability to recognize quality before consensus has formed around it. In music, the producer who can generate endless alternatives still has to know when the chorus is emotionally true, when an arrangement is crowded, and when another revision will make the song worse. In strategy, the same pattern appears as disciplined choice among plausible options.
The stopping problem
Abundant generation also creates a stopping problem. When another version costs almost nothing, why stop? The answer cannot come from the machine alone because the machine is optimized to continue responding. Completion requires a human standard. Good selectors know when the marginal option no longer improves the decision. They protect attention from abundance.
Taste after abundance
Selection becomes economically important when the cost of generating another option approaches zero. In the old environment, producing ten credible alternatives might have been expensive. In the new environment, the expensive act is deciding which one deserves attention.
Music makes the distinction easy to hear. A system can produce variations rapidly: different tempos, textures, harmonic choices, lyrical directions, arrangements, and mixes. More options do not automatically produce a better record. At some point someone has to say this is the one, this is not, this line is emotionally false, this arrangement is technically impressive but wrong for the artist, this version serves the scene.
The same problem appears in strategy. AI can generate twenty priorities for an institution. A serious strategy cannot contain twenty priorities. Leadership has to choose what will receive money, time, political capital,