Productive for Being Productive: What Paris Hilton and LLMs Have in Common

What do Paris Hilton and Large Language Models Have in Common?

Paris Hilton was famous for being famous, which set up a self-referential loop where visibility generated visibility, and since the metrics said she mattered, she did.

I’m starting to think AI is productive for being productive.

In The Wall Street Journal this week: 40% of executives say AI saves them more than eight hours a week, while two-thirds of their employees say it saves them two hours or less. How do the same tools in the same companies create such opposite realities?

Workday found the mechanism: time “saved” by AI gets eaten by correcting errors and reworking AI-generated content. Because executives are generating, their subordinates are fixing, and work isn’t moving faster as much as it is sliding downhill faster. Workslop, they call it.

Meanwhile, 56% of CEOs say AI has delivered no financial benefit, though the dashboards look great and everyone’s producing more of, well, something?

IDK if this is very different from philanthropy/social change, where individual interventions can look impressive but leave the underlying system untouched. More of a feeling/vibe of progress decoupled from the fact of it.

So long as we measure outputs, not outcomes, count what individuals produce, not what organisations accomplish, the gap between those is where collective capacity will go to die.

The thing is, Paris Hilton was in on the joke and built real businesses behind the performance.

Originally written for LinkedIn on 23 January 2026. View original →