Hierarchy as Information Relay

The reframe

Jack Dorsey's starting observation is that management hierarchy is a roughly two-thousand-year-old structure, borrowed from military and religious organizations and iterated on slowly ever since, whose function is information relay: making communication manageable across a large population. A layer of vice presidents tells the layer of directors what leadership decided; the directors synthesize what their teams produced and report it back upward. The hierarchy, in this framing, is the communication network itself.¹

The counterintuitive part of the claim is that hierarchy was never primarily about distributing power. Power and accountability were downstream effects of the communication function: the hierarchy existed to route signal, and authority accreted around the people sitting at the routing nodes because they were the ones with the most information, not because the structure was designed to grant them authority in the first place.

What happens when relay gets cheap

If hierarchy is fundamentally an information-relay technology, the interesting question becomes what happens once relaying information gets dramatically cheaper. Every artifact a company already generates, chat messages, emails, pull requests, code, recordings, documents, is a structured information stream about what the company is actually doing. An intelligence layer that ingests all of it can make any part of that information queryable by anyone without a lossy human relay in between, surface patterns and anomalies without a manager needing to notice and report them, and generate board-level context without a week of deck-building. What remains of management that this layer cannot replace is judgment, taste, and creativity on one side, ownership and accountability on another, and coaching through doing on a third, the durable human functions once the relay itself is automated.

Why the shift is different this time

Hierarchy has been embedded in organizational design for so long that it is nearly invisible as a technology choice rather than a law of nature. The economic version of the same idea is that a firm grows only until its internal coordination costs exceed the benefit of staying inside it, and that coordination cost is fundamentally the cost of relaying information between people who need to act coherently. The internet and email were the last major pressure on this system, and they lowered the cost of communication without eliminating the need for relay intermediaries altogether. AI, on this account, is the first technology to eliminate the relay function outright rather than merely making it cheaper.

The practical destination Dorsey describes is a company that is entirely legible, every aspect of it visible to the intelligence layer and therefore queryable by anyone in it: "Imagine if your company was entirely legible, every aspect of it. We're not far off from that from a data perspective."¹ The data already exists, since every digital action inside a company creates an artifact; what has been missing until now is the intelligence layer that makes all of it coherent and queryable at once, which is what makes flat reporting structures tractable even at large scale. Block's own reporting depth illustrates the scale directly: a maximum depth of five layers, a target of two to three, and an ideal of zero, all six thousand employees reporting to the chief executive directly, mediated by the intelligence layer.¹

Open questions

Relay and judgment are not fully separable in practice: middle management does not just pass information along, it also filters and contextualizes what counts as signal versus noise before anything moves upward, so if an intelligence layer inherits that filtering role, the quality of its resulting picture of the company depends entirely on how well those filters are designed. Information relay also historically created accountability, since a manager who relayed something upward took responsibility for its accuracy; who is accountable for what an intelligence layer reports is a genuinely open question. And making a company fully legible means every failure and every deviation becomes visible, which risks people adapting their behavior to satisfy the model rather than to fit the actual situation on the ground, the hierarchy's old opacity having sometimes been a feature rather than a bug.