You Flattened Management. What Did You Actually Remove?
Cutting a management layer removes three bundled jobs at once — and AI can only take one of them. Decompose the role before you compress it.
Nearly half of US companies stripped out a layer of management in the past year, and the pitch was always the same: flatter, leaner, faster, powered by AI. If work has felt subtly broken since, you are not imagining it. Something load-bearing left with the org chart.
Management is three jobs, not one
- Routing — information logistics, aggregating up and cascading down. This is the best-documented job, and the one AI genuinely takes. If a meeting could have been an email, this is what failed.
- Sensemaking — pulling signal out of noise and deciding what actually matters. It draws on years of domain context, and AI is, at best, a partner here.
- Accountability — ownership and feedback, the felt sense of being answerable for an outcome over time. Agents don't run long enough to own anything the way a person does.
Companies "automate management" by seeing the first job and quietly discarding the other two.
Three live experiments
One flat model concentrates sensemaking in a handful of founders and leaves accountability to self-reflection — fast, but with real human casualties. Another decomposes deliberately: software handles routing, time-boxed owners handle sensemaking, and player-coaches keep a human in the accountability loop. A third simply compresses — fewer managers, wider spans, sharper consequences.
Decompose before you compress
The firms that hold up won't be the ones doing more with fewer managers. They'll be the ones who took the role apart, kept what's human, and could specifically imagine where AI fits. Removing a layer without naming its jobs isn't a strategy — it's a crisis of imagination.