Lead: While most companies still view AI as a mere "efficiency tool," Jack Dorsey pushes the question further: Is AI fundamentally rewriting the operational logic of organizations themselves? As co-founder and former CEO of X and founder of Block, he has long explored the relationship between technology and organizational form.

This article traces historical evolution to re-examine why enterprises have taken their current shape—and why this structure is now beginning to unravel. From Roman legions to modern corporations, two millennia of organizational development have been constrained by a single limit: transmitting information and coordinating action within finite span of control. Hierarchical structures, middle management, matrix systems—these are all different solutions to the same underlying problem.

AI’s emergence now challenges this foundational assumption for the first time. When information can be modeled, understood, and distributed in real time, does an organization still require human-centered coordination mechanisms?

Such shifts are already manifesting in practice. Recently, a report by Renwu revealed that the AI company “Moonshot” operates with over 300 employees across a structure devoid of departments, titles, or OKRs/KPIs; collaboration relies on direct communication rather than hierarchical reporting. Five co-founders each directly interface with 40–50 employees. Meanwhile, agents are embedded into daily workflows, capable of rapidly handling information aggregation, product design, and even code generation. This structure isn't simply "de-managing"—it proactively relocates complexity into hiring, talent mobility, and tooling systems.

Using Block’s own practices as a lens, this article advances a more radical vision: from hierarchical organizations to intelligent companies, replacing traditional information routing with a company world model + customer world model + intelligent layer, potentially redefining middle management itself. This is not merely about efficiency—it may represent a fundamental rewrite of organizational form.

Below is the original text:

In Sequoia Capital’s view, "velocity" is the best predictor of a startup’s success. Most companies still see AI as a productivity booster. Only a few are beginning to examine how AI transforms collaboration. Block is demonstrating a new path: fundamentally reconstructing organizational design and leveraging AI as a compounding, sustainable competitive advantage that continuously amplifies speed.

Two millennia before organizational charts emerged, the Roman army had already solved a problem that still plagues large-scale organizations today: how to coordinate tens of thousands of people under conditions of limited communication and vast distances.

Their solution was a nested command hierarchy, maintaining relatively stable span of control at each level. The smallest unit was the contubernium, consisting of eight soldiers sharing a tent, equipment, and a donkey, led by a decanus. Ten contubernia formed a centuria (approximately 80 men), commanded by a centurion. Six centuria composed a cohors; ten cohors made up a legion of roughly 5,000 soldiers.

At every level, clear commanders aggregated information upward and issued directives downward. This progression—from 8 → 80 → 480 → 5,000—was essentially an efficient information transmission mechanism built upon a simple yet critical premise: one person can effectively manage only about 3 to 8 others. The Romans gradually discovered this principle through prolonged warfare. Even today, the U.S. military’s structure largely follows similar logic. We call this constraint the span of control, which remains an inescapable foundational limit for all large organizations.

The next major transformation came from Prussia.

After their devastating defeat at the Battle of Jena in 1806, Scharnhorst and Gneisenau led sweeping military reforms, proposing an uncomfortable truth: success could no longer rely on individual genius but must depend on systems. They established the General Staff, cultivating a cadre of professional officers whose role was not combat, but planning operations, processing information, and cross-unit coordination. Scharnhorst’s goal was to "compensate for incompetent generals by providing capabilities they lacked." This was, in essence, the prototype of middle management: a group of specialists responsible for information flow, pre-emptive decision modeling, and sustaining coordination within complex organizations. At the same time, the military clearly differentiated between line (executing core missions) and staff (providing expert support)—a distinction still widely adopted by corporations.

In the 1840s and 1850s, American railroads imported the military’s hierarchical framework into the commercial world.

The U.S. Army supplied numerous engineers trained at West Point, who brought military organizational thinking into industry. The line-and-staff structure, business unit divisions, and bureaucratic reporting and control systems were originally born in the military. In the mid-1850s, Daniel McCallum of the New York & Erie Railroad drew the world’s first organizational chart to manage a 500-mile railway system and thousands of employees. Informal management methods used for smaller railways had proven inadequate, leading to frequent train collisions. McCallum institutionalized Roman-style hierarchy: clear authority levels, explicit reporting lines, and structured information flows. This became the archetype of the modern corporation.

Subsequently, Frederick Taylor ("the father of scientific management") optimized this system internally. He broke work into specialized tasks assigned to trained experts, managed through quantifiable metrics rather than intuition, thus forming the functional pyramid—an organizational form maximizing efficiency within existing information routing systems.

The first major stress test for this functional structure came during the Manhattan Project in World War II. The project required interdisciplinary collaboration among physicists, chemists, engineers, metallurgists, and military personnel, achieving a singular goal under extreme secrecy and time pressure. Robert Oppenheimer employed functional divisions at Los Alamos but insisted on open cross-departmental collaboration, resisting military attempts to isolate teams. In 1944, when the "implosion problem" became a critical bottleneck, he restructured teams into cross-functional units—an unprecedented move in corporate practice at the time. This model succeeded, but it was a wartime exception driven by a few exceptional individuals. The postwar question for businesses was whether such cross-functional collaboration could be sustained.

Postwar expansion in scale and globalization made the limitations of functional structures increasingly evident.

In 1959, Gilbert Clee and Alfred di Scipio of McKinsey published "Creating a World Enterprise" in the Harvard Business Review, proposing the matrix organization framework—integrating functional expertise with business unit structures. Under Marvin Bower’s leadership, McKinsey helped Shell, GE, and others implement this model, balancing "central standards" with "local flexibility." This system became the dominant paradigm of the modern enterprise in the postwar global economy.

Then, to address the complexity and bureaucratization inherent in matrix structures, new management frameworks continually emerged.

In the 1970s, McKinsey introduced the 7-S Model, distinguishing "hard elements" (strategy, structure, systems) from "soft elements" (shared values, skills, people, style), emphasizing that organizational effectiveness cannot be ensured by structure alone—culture and human dynamics are equally essential.

Entering recent decades, tech companies have conducted bolder experiments in organizational design.

Spotify introduced cross-functional squads and short-cycle iterations; Zappos experimented with Holacracy, eliminating managerial titles; Valve adopted a flat structure with no formal hierarchy. These attempts exposed the limitations of traditional hierarchies, but none fully resolved the core issue: Spotify reverted to conventional management as scale grew, Zappos faced massive attrition, and Valve’s model proved unscalable beyond a few hundred people. When organizations reach thousands of employees, they inevitably return to hierarchical coordination—because no more effective information routing mechanism exists.

This constraint mirrors exactly what the Romans and WWII Marines faced: shrinking span of control forces more layers, and increased layers slow down information flow. For two thousand years, organizational innovation has attempted to circumvent this trade-off—but never truly broken it.

At Block, we begin to question a foundational assumption: that organizations must use humans as coordinators via hierarchical structures. Our goal is to replace the function of hierarchy with systems. Most companies today merely equip employees with AI co-pilots, making existing structures slightly more efficient—without changing their essence. We aim to build something else entirely: a company itself becomes an agent (even a miniature AGI).

We are not the first to attempt transcending hierarchy. Haier’s "people-single integration," platform-based organizations, and data-driven management are similar explorations. But they lack a crucial element: a technology capable of genuinely assuming coordination functions. AI is that technology. For the first time, a system exists that can continuously maintain a holistic model of enterprise operation and coordinate based on it—without requiring human intermediaries to relay information through layers.

To achieve this, a company needs two things: a world model of its own operations, and sufficiently rich customer signals.

Block operates primarily remotely, so all work leaves traceable digital "footprints": decisions, discussions, code, designs, plans, issues, and progress. These constitute the raw material for the company’s world model.

In traditional firms, managers interpret team status and relay information up and down. In a machine-readable organization, AI can continuously construct this global view: what’s being done, where bottlenecks occur, how resources are allocated, what’s effective, and what isn’t. These insights, once carried by hierarchy, are now held by models.

But system capability depends on input signal quality—and money is the most truthful signal. People may lie in surveys, ignore ads, abandon shopping carts—but when they spend, save, transfer, borrow, or repay, those actions are real. Block sees both sides of every transaction daily: buyers via Cash App, sellers via Square, and merchant operating data. This enables the construction of a rare customer world model—one grounded in authentic transaction signals, enabling granular, real-time understanding of individual customers and merchants, with signals that accumulate and strengthen over time.

The company world model and customer world model together form the foundation of a new kind of company. In this model, the company no longer operates around fixed roadmaps defined by product teams—but instead centers on four core pillars:

First, capabilities: foundational financial functions like payments, lending, card issuance, banking, buy-now-pay-later, payroll—these are not products, but underlying modules without user interfaces, yet subject to reliability, compliance, and performance requirements.

Second, world models: including the company model (understanding internal operations) and customer model (representations of customers and markets derived from transaction data), evolving toward systems with causal reasoning and predictive power.

Third, the intelligence layer: actively composing capabilities for specific customer combinations at precise moments. For example, when the system predicts a restaurant’s cash flow will tighten, it automatically bundles loan and repayment options and proactively sends them; or when user behavior hints at a move, it auto-configures a new suite of financial services. None of these require prior product design by a product manager.

Fourth, interfaces: such as Square, Cash App, Afterpay, TIDAL—these are merely delivery surfaces; true value emerges from the model and intelligence layer.

When the system attempts to compose a solution but lacks a certain capability, that "failure signal" becomes the future product roadmap. The traditional process of product managers hypothesizing demand is replaced directly by real customer behavior.

In this structure, the organization itself transforms. In traditional firms, intelligence resides in people and is routed through hierarchy. Here, intelligence lives in the system, and humans occupy the edge. The edge is where intelligence meets reality. Humans can sense intuitions, cultural nuances, trust dynamics, and complex contexts that models miss—and play critical roles in ethics and high-risk decisions. Yet they no longer need to coordinate through hierarchy, because the world model provides the necessary context.

In practice, the organization simplifies into three roles:

· IC (Individual Contributor): experts building capabilities, models, and interfaces;

· DRI (Directly Responsible Individual): mobilizing resources around specific problems or customer outcomes;

· Player-coach: simultaneously engaged in frontline work and talent development, replacing traditional managers.

No fixed middle management layer is needed—the rest of coordination is handled by the system.

Block is still in the early stages of this transition—a difficult journey, with some experiments likely to fail. But we share this direction because we believe every company will eventually face the same question: Are you continually deepening your understanding of a complex problem?

If the answer is no, AI is just a cost-reduction tool. If yes, AI reveals the company’s true essence.

Block’s answer is the economic graph: connecting millions of merchants and consumers, continuously understanding behavior on both sides of transactions, and accumulating knowledge in real time. We believe this model—organizing companies through intelligence rather than hierarchy—will reshape how all kinds of enterprises operate in the coming years.

The speed of a company ultimately hinges on the velocity of information flow. Hierarchy and middle management slow this flow. For two thousand years, from Roman legions to modern enterprises, we had no better alternative. Now, that premise is changing. Block is building the next form.