When Making Things Is Free, Deciding Becomes the Job

AI collapses the cost of producing options. It does nothing to collapse the cost of choosing between them. That gap is where organizations will quietly break — and where the next scarce skill is forming.

I keep watching the same meeting happen at different companies.

A team that used to bring one prototype to review now brings nine. All of them are decent. The AI made it trivially cheap to build all nine, so why wouldn't you? The meeting runs long. Nobody can articulate why one direction beats the others. The decision gets deferred, a tenth variant gets commissioned as a compromise, and everyone leaves feeling productive.

Nothing shipped. But an enormous amount got made.

That meeting is a preview of the next organizational disease, and it comes straight out of the second-order playbook: make something abundant, and the scarcity moves next door.

Scarcity was doing your strategy for you

Here is what expensive execution actually did for organizations, invisibly, for decades: it forced prioritization.

When building a prototype took a quarter and three engineers, you could not try everything. So you argued first. The arguing was annoying, but it was doing real work — it was the mechanism by which organizations decided what they believed. Constraint was a filter. Budgets were an editing function. The cost of making things was quietly running the strategy process, and nobody had to be good at deciding because scarcity decided for them.

Cheap execution removes the filter. Every half-formed idea can now become a working artifact by Friday. The org's production capacity has been multiplied by ten; its decision capacity — the attention of the people who must evaluate, compare, and commit — has been multiplied by exactly one.

The result is a traffic jam in front of human judgment. Work-in-progress piles up. Everything is "promising." The backlog of plausible things grows without bound, and the organization mistakes the size of the pile for the rate of progress.

Producing more was never the constraint

The uncomfortable truth cheap execution exposes is that most organizations were never actually bottlenecked on production. They were bottlenecked on conviction — on the ability to decide what they wanted and hold that decision under pressure. Expensive execution just made the conviction problem look like a resourcing problem.

AI strips the disguise off. When anyone can generate ten strategies, ten designs, and ten roadmaps before lunch, the differentiating question stops being "can we build it?" and becomes "should this exist, and who is willing to be wrong about it?"

Notice what that second question requires. Not intelligence — the models have plenty. It requires wanting something. A model can rank options against criteria, but someone has to supply the criteria, and the criteria are the strategy. Taste, in the serious sense: a considered view of what good looks like, held consistently enough that a thousand cheap options can be filtered against it quickly.

Abundance doesn't kill strategy. It reveals which organizations never had one.

What the new scarce work looks like

Follow the bottleneck and you can see the labor market forming around it, the way it always does.

The editor becomes the power role. Not editor of text — editor of portfolios. The person who can look at forty AI-generated candidates and kill thirty-eight of them fast, with reasons, becomes more valuable than the person who generated them. Killing options is emotionally expensive and reputationally risky, which is exactly why it will command a premium. Every abundant-content industry already learned this: when anyone can make things, the scarce person is the one who decides what's worth making. That logic is now coming for strategy decks, product roadmaps, and org designs.

Decision hygiene becomes infrastructure. Organizations will need explicit systems for what used to happen implicitly: who decides, on what criteria, by when, with what kill-rate. The companies that treat deciding as a designed process — with throughput, quality control, and postmortems — will metabolize AI abundance. The ones that route every option through the same overloaded senior brains will experience AI as a productivity tool that somehow made everything slower.

Accountability becomes the human moat. When the analysis is machine-made and every option comes with a fluent justification, the thing that cannot be delegated is ownership of the outcome. Someone has to put their name on the choice. I made a version of this argument about trust — production gets cheap, confidence gets expensive — and it applies with full force inside organizations: the signature is the scarce good.

The trap: deciding faster instead of choosing better

There is a seductive wrong answer to all this, and plenty of companies will take it: use AI to decide faster too. Auto-rank the options. Score the portfolio. Let the model pick.

Some of that is fine for reversible, low-stakes calls. But notice what it does at the top of the funnel: it turns the decision process into more production. The ranking is one more artifact. The scores are one more input somebody has to judge. You have not escaped the bottleneck; you have added a very articulate lobbyist for every option in the pile.

The organizations that win will do something less comfortable. They will choose less, better. Fewer bets, made with more conviction, killed more decisively when wrong. They will treat their decision bandwidth as the fixed asset it is and spend it deliberately, the way they once spent scarce engineering time. Strategy used to mean deciding what you could afford to try. Now everything is affordable to try.

Strategy is becoming the discipline of deciding what deserves to exist.

The series, in one line

This is the fourth essay in a row circling the same mechanism, so it is worth saying plainly: every time AI makes something abundant — expertise, cognition, execution — the value doesn't vanish. It moves to whatever the abundance newly strains: judgment formation, physical capacity, and now the deciding mind itself.

First-order thinking asks what the machine can make. Second-order thinking asks what all that making will demand of us.

The answer, increasingly, is the oldest executive function there is: knowing what you want.