“Flat” isn’t the goal. Deliberate modularity is.
Every few years someone announces the end of hierarchy. Go flat, remove the managers, let self-organization do the rest. It rarely survives contact with reality. Within a year the “flat” organization has quietly grown a new informal hierarchy, or it has fragmented into teams that no longer pull in the same direction. The flat-org dream keeps failing for the same reason: it confuses removing structure with changing structure.
There’s a more durable design hiding in the cybernetic tradition, and it has a less exciting name: deliberate modularity. The idea is to build the organization out of autonomous, value-creating units — each with the authority to act on the information in front of it — and then connect them with just enough shared machinery to keep them coherent. Not flatness. Not central command. A structured network.
What holds it together is a small set of functions, each doing a specific job:
- Autonomous units that actually deliver value and have real authority to decide locally, because that’s where the relevant information arrives first.
- Shared protocols and rhythms — the coordination layer that lets independent units mesh without anyone phoning the centre. This is the part flat-org experiments forget, and it’s why they re-centralize under stress.
- Internal audit and resourcing — a way for the whole to keep its units healthy and accountable without micromanaging them.
- Outward scanning — someone whose job is to watch the changing environment and feed that back in.
- A clear identity and policy — a shared answer to who we are and what we will and won’t do, so units can act independently without drifting apart.
The reason this matters now is economic. Bureaucracy was a sensible response to a slow world with expensive information. When change was gradual and communication costly, centralizing decisions — funnelling variety up to a few people at the top — was efficient. But as change accelerates and information gets cheap, that central funnel becomes the bottleneck. The volume and speed of what the organization has to respond to exceeds what any centre can process. The only way to keep up is to handle that variety at the edge, where it actually shows up.
The elegant part is the mechanism that prevents fragmentation: recursion. Each autonomous unit is itself organized the same way — it has its own version of doing-the-work, coordinating, sensing, and holding identity. The pattern repeats at every scale, like nested dolls. That’s what lets you grant edge autonomy without the organization flying apart: every level is self-regulating, and every level shares the same logic, so coherence comes from the design rather than from a person at the top enforcing it.
Now the honest part. This design is strong on motive and weak on discipline. It’s easy to be persuaded that autonomy is good; it’s hard to install the unglamorous coordination machinery — the protocols, the shared rhythms, the rules of recursion — that keeps a network from quietly re-centralizing the moment things get tense. Under pressure, autonomy is the first thing organizations claw back. And the blueprint inherits a real unsolved problem: accountability. When authority is genuinely distributed, who is answerable when something goes wrong — legally, financially, reputationally? “The network decided” is not an answer a regulator accepts.
So the takeaway isn’t “go flat.” It’s the opposite: autonomy at the edge only works if you over-invest in the coordination and identity that hold it together. The interesting design question is never how to remove structure. It’s how to build the right structure — modular, recursive, and deliberate — so that distributed speed doesn’t cost you coherence.
Next week: even a perfect structure fails if signals can’t travel through it. The problem of the idea that dies in someone’s inbox.
Part of the countdown to Metaphorum 2026 — “100 Years of Stafford Beer” (17–19 September 2026, Alliance Manchester Business School, Manchester), where I’m convening a half-day “Cybernetic Futures” workshop [workshop slot — TBC]. Programme, themes, and registration: conference2026.metaphorum.org. Abstract deadline: 17 July 2026.