When Intelligence Replaces the Org Chart

Sequoia and Block just published the clearest articulation of something I've been trying to build for a year. The thesis is one sentence: hierarchy exists to route information, and AI can now route it better. Everything else follows from that.

It's worth understanding how old the constraint is before you appreciate how big the claim is.

Two thousand years of the same problem

The Roman Army figured out something every company still runs on: a single leader can only manage three to eight people. So you nest hierarchies. Eight soldiers in a tent under a decanus. Ten tents make a century of eighty under a centurion. Six centuries make a cohort, ten cohorts a legion of roughly five thousand. The structure — 8 to 80 to 480 to 5,000 — was an information routing protocol built around one human limitation. We now call it span of control, and it still governs every large organization on earth.

The Prussians added a General Staff after Napoleon crushed them at Jena: professional officers whose job was not to fight but to process information and keep a complex force aligned. That was middle management before the word existed. The American railroads put it on an org chart in the 1850s, when informal management started killing people in train collisions. Taylor optimized the pyramid. The matrix came after the war. Then Spotify squads, Holacracy, Valve's flat org.

Every experiment hit the same wall. As companies scaled into the thousands, they reverted to hierarchy. Not because people love middle management — because no other mechanism for routing information existed. Narrower span of control means more layers; more layers means slower information flow. Two thousand years of org design has been an attempt to work around that tradeoff without breaking it.

What's actually different now

Most companies using AI today hand everyone a copilot. That makes the existing structure work slightly better without changing it. Block is doing something else: replacing what the hierarchy itself does. The company becomes an intelligence.

For that to work you need two things. A world model — a continuously updated picture of how the company understands itself and its customers. And a customer signal rich enough to make that model useful. Block has both. It's remote-first, so every decision, design, and line of code already exists as a machine-readable artifact — raw material for a company world model. And it sees both sides of millions of transactions a day through Cash App and Square. Money is the most honest signal there is. People lie on surveys and abandon carts, but when they spend, send, borrow, or repay, that's the truth.

Four layers, three roles

Instead of product teams building predetermined roadmaps, the company builds four things. Capabilities — the atomic primitives like payments, lending, and card issuance, with no UI of their own. A world model — how the company understands itself and its customers. An intelligence layer — which composes capabilities into solutions for specific moments and delivers them proactively, before a customer thinks to ask. And interfaces — Square, Cash App, Afterpay. Delivery surfaces, not where the value lives.

The clever part: when the intelligence layer tries to compose a solution and can't because a capability is missing, that failure is the roadmap. Customer reality generates the backlog directly, instead of a product manager guessing what to build next.

And the people? Three roles. Individual contributors who build and operate the system. DRIs who own a cross-cutting problem and a customer outcome, with authority to pull resources across teams. And player-coaches who combine craft with growing people — they still build, they just don't spend their days in status meetings, because the world model handles alignment. No permanent middle management. The edge is where the action is, and the model gives every person at the edge the context a manager used to provide.

The same bet, named

This is the bet we've been making at Thrivbe. We don't have a CEO; we have a network. Decisions happen closest to the problem. The boss is the mission, not a person. We've been calling it distributed leadership. Block is naming the deeper shift underneath it: the hierarchy was always just an information routing system, and humans were never the only option for the job.

One line in the essay stopped me cold: “What does your company understand that is genuinely hard to understand, and is that understanding getting deeper every day?”

If the answer is nothing, AI is a cost-cutting story. You trim headcount for a few quarters and get absorbed by something smarter. If the answer is deep, AI doesn't augment your company — it reveals what your company actually is.

Eight soldiers in a tent needed a decanus. Eighty needed a centurion. Five thousand needed a legate. The question was never whether you needed layers. It was whether humans were the only thing that could do what those layers do. They aren't anymore.