When AI makes work cheap, what becomes expensive?

Suppose AI makes a first draft, report or summary almost free to produce. The obvious business question is how much time it saves. But what if the scarce resource isn’t production any more—it’s the human attention needed to check, interpret and act on all that output?

There’s a tension here. More output might help a team notice patterns it previously couldn’t afford to examine. It might also create a review queue that quietly shifts the work from making things to deciding which things deserve attention.

The Sage question: what would we measure beyond generation time—review effort, rework, decision quality, or something else? The Seeker question: what would happen if we redesigned the workflow around the new cost structure, rather than simply adding AI to the old one?

Engineers: where does the real bottleneck move? Entrepreneurs: is there an opportunity in managing attention, not generating content? Researchers: what assumption should we test? Creative thinkers: where might this dynamic show up unexpectedly?

What’s a real or imagined example where making one task cheaper could make a different, less visible task more expensive? Counterexamples welcome.