A product team can now produce more in one afternoon than it used to produce in a week: landing page variants, onboarding flows, support copy, internal dashboards, pitch decks, small prototypes that look almost real. This is useful. It is also slightly dangerous, because movement has always been easy to confuse with progress, and AI makes that confusion cheaper, faster, and much harder to notice from the outside.
The bottleneck moved
For a long time, the obvious constraint in digital work was production. Could we design it, write it, build it, test it, translate it, ship it? That constraint has not disappeared, but in many areas it has become less decisive. The uncomfortable part is that when production becomes easier, another weakness becomes visible: the ability to choose. Not to choose between two button colors, although that also happens, but to choose what deserves attention, what should be killed early, what is coherent with the promise of the product, and what is merely a well-rendered distraction.
This is not an argument against AI-assisted work. That would be lazy. The tools are useful, and in many cases they remove pointless friction. The point is narrower and, in my opinion, more important: when the cost of making things drops, the cost of weak selection rises. A team that could once hide behind limited capacity now has to reveal whether it has a real standard, or only a backlog with better lighting.
Taste is not decoration
I know some people dislike the word taste, especially in business, because it can sound elitist or vague. That reaction is understandable. Badly used, taste becomes a way to end discussion without evidence. Someone senior says something “feels off”, and everyone else is supposed to translate the mood into work. That is not taste. That is power wearing a slightly softer jacket.
Product taste, used responsibly, is something else. It is calibrated sensitivity to what is useful, proportionate, coherent, and humane in a given context. It is the ability to notice that a feature technically solves a problem but emotionally makes the product feel needy. It is noticing that a dashboard answers a question nobody important is actually asking. It is understanding that more personalization can sometimes reduce trust, because the user starts wondering who is really in control of the interaction.
None of this replaces research, data, strategy, or technical skill. It connects them. Taste is not a mystical gift reserved for designers or founders with strong opinions. It is a trained ability to keep several forms of evidence in view without pretending that any single one of them can make the decision for you.
Abundance can protect mediocrity
The strange thing about abundant output is that it can make a weak team look disciplined for a while. There are more artifacts, more options, more versions, more notes after meetings, more screenshots in Slack, more “directions” to compare. It looks like seriousness. Sometimes it is. Often it is a socially acceptable way to postpone a harder conversation: what are we actually trying to make true for the customer, and what are we willing to ignore?
This is where mediocre work becomes more resilient, not less. If producing another variant is cheap, the team can keep the decision alive almost indefinitely. The debate moves from substance to comparison. People stop asking whether the idea is strong and start asking which generated version is strongest. That difference is subtle, but it matters. The first question is about reality. The second can become an internal game with attractive screenshots.
Leadership becomes editing
In this environment, leadership is less about approving more work and more about editing the shape of attention. I do not mean editing as polishing copy. I mean the willingness to reduce, reject, combine, delay, and sometimes say that a perfectly competent output does not belong in the product. This can feel uncomfortable, because it makes judgment visible. It also makes accountability visible. When a leader says yes to everything because the team can now produce everything, that is not openness. It is a refusal to impose a meaningful standard.
There is a useful discipline here: before asking how fast something can be generated, it may be worth asking what ability the generation will amplify. If the team has clarity, AI can make that clarity travel faster. If the team has confusion, AI can give confusion a cleaner interface. The tool does not care. The organization should.
The real advantage is restraint
Speed still matters. In competitive markets, pretending otherwise would be naive. But speed is not a substitute for taste, and it becomes actively harmful when it helps a team avoid developing one. The teams worth watching will probably not be the ones producing the most artifacts per week. They will be the ones that can absorb new capability without losing their sense of proportion.
That is less glamorous than another demo. It is also harder to fake. And maybe that is the point: when making gets easier, choosing becomes the work. The rest is just output.