Take any DTC/ecommerce startup and you'll find a familiar loop: grow revenue, keep LTV/CAC and payback periods tolerable. Marketing dollars buy traffic, traffic converts into orders, orders fund more marketing.

Within the loop, there are three operating layers:

• Judgment: calling the plays, making the bets, managing expectations, owning the stakes
• Execution: running the campaigns, building the features, retaining customers, doing the work
• Data: measuring outcomes, canning reports, justifying narratives, existing

AI is eating the bottom two. More data, more doing, more agentic.

But then AI smashes into a human core: judgment — born out of tacit knowledge, institutional knowledge, domain expertise. It requires a face and a neck to carry the stakes, a physicality that AI can't simply devour. Yelling at Claude when the brand campaign misses just doesn't offer the same level of catharsis.

And judgment leaves no artifacts. Execution leaves commits, receipts, post-mortems. Data leaves dashboards, semantic layers, revenue forecasts. Judgment that dictates the other two leaves a calendar invite and a vibe.

So what is judgment, and why does it leave no trace? In organizations, it's navigating a never-ending, multiplayer prisoner's dilemma: cooperation benefits everyone but self-preservation is the safe bet. So nobody shows their hand.

This means picking your battles, knowing which ones to lose on purpose. And the most important calls are calls, never Slack.

AI is transferring the bottleneck from the execution layer up to the judgment layer. With humans remaining in the loop, will the pace of AI exacerbate theater or eliminate the need for it?

One thing's for sure — the future of work...is more work.