Rethinking Startup Hierarchy in the AI Era
Jack Dorsey argues that most companies are still built on an old constraint: humans have historically been the only reliable way to route information across growing organizations. In his words, “Companies move fast or slow based on information flow.” For centuries, hierarchy solved that problem. Managers gathered context, passed signals upward, translated decisions downward, and kept teams aligned.
The AI era challenges that assumption.
If software can continuously synthesize what is being built, what is blocked, how customers are behaving, where resources are going, and which priorities matter most, then part of the traditional management layer stops being structurally necessary. Dorsey’s point is not that people stop mattering. It is that much of what organizations call management has really been information logistics. Once machines can handle more of that work, startups can redesign themselves around contribution, ownership, and development rather than reporting chains.
Dorsey and Roelof Botha describe this as moving from hierarchy to intelligence. They are not talking about giving everyone a chatbot and calling it transformation. They are describing something more radical: a company that builds an internal “world model” of itself and pairs it with a deep model of customers and markets. In their framing, AI becomes part of the operating system of the company, not just a productivity feature.
That leads to a different pattern of work.
The first role is the individual contributor: people whose time is spent building, designing, selling, coding, operating, or solving. In an AI-native organization, they should receive context directly from the system instead of waiting for it to pass through several human layers. The second role is the Directly Responsible Individual, the person who owns a specific outcome and has the authority to pull the necessary resources together across functions. The third role is the player-coach: someone who still does the work but also develops people, raises standards, and helps others navigate complexity. This is not the old manager as coordinator and status collector. It is a practitioner-leader focused on craft and growth.
This model has obvious appeal for startups.
It can reduce approval friction. It can move decisions closer to customers. It can protect makers from coordination overload. It can expose reality faster by replacing filtered updates with direct visibility into what is happening across the company. For fast-moving teams, that can become a real competitive advantage.
But the thesis should not be romanticized.
Not all management is bureaucracy. Good managers do more than move information. They resolve conflict, build judgment, protect culture, sequence work under uncertainty, and make calls when signals are incomplete or contradictory. AI can support those tasks, but it does not remove their human weight. A startup that eliminates management without building clarity, trust, and decision discipline will not become more intelligent. It will become more chaotic.
That is the real test.
The winners in the AI era will not be the companies that simply cut layers. They will be the ones that redesign work around clearer accountability, richer operating visibility, and stronger human judgment. AI can flatten information flow. It cannot substitute for ethics, trust, conviction, or taste.
So the future is probably not a manager-free startup. It is a startup with fewer people doing coordination for coordination’s sake, more people directly accountable for outcomes, and a much smaller gap between what the company knows and what it does.
That is the deepest implication of Dorsey’s argument. AI should not just help startups work faster inside old structures. It should force founders to ask whether those structures still deserve to exist at all.
The most important takeaway is simple: the org chart of the AI era should be built around contribution, ownership, and development, while machines increasingly handle routing, synthesis, and visibility. Startups that understand this early may not just become more efficient. They may become structurally better at learning, deciding, and adapting.
Jack Dorsey argues that most companies are still built on an old constraint: humans have historically been the only reliable way to route information across growing organizations. In his words, “Companies move fast or slow based on information flow.” For centuries, hierarchy solved that problem. Managers gathered context, passed signals upward, translated decisions downward, and kept teams aligned.
The AI era challenges that assumption.
If software can continuously synthesize what is being built, what is blocked, how customers are behaving, where resources are going, and which priorities matter most, then part of the traditional management layer stops being structurally necessary. Dorsey’s point is not that people stop mattering. It is that much of what organizations call management has really been information logistics. Once machines can handle more of that work, startups can redesign themselves around contribution, ownership, and development rather than reporting chains.
Dorsey and Roelof Botha describe this as moving from hierarchy to intelligence. They are not talking about giving everyone a chatbot and calling it transformation. They are describing something more radical: a company that builds an internal “world model” of itself and pairs it with a deep model of customers and markets. In their framing, AI becomes part of the operating system of the company, not just a productivity feature.
That leads to a different pattern of work.
The first role is the individual contributor: people whose time is spent building, designing, selling, coding, operating, or solving. In an AI-native organization, they should receive context directly from the system instead of waiting for it to pass through several human layers. The second role is the Directly Responsible Individual, the person who owns a specific outcome and has the authority to pull the necessary resources together across functions. The third role is the player-coach: someone who still does the work but also develops people, raises standards, and helps others navigate complexity. This is not the old manager as coordinator and status collector. It is a practitioner-leader focused on craft and growth.
This model has obvious appeal for startups.
It can reduce approval friction. It can move decisions closer to customers. It can protect makers from coordination overload. It can expose reality faster by replacing filtered updates with direct visibility into what is happening across the company. For fast-moving teams, that can become a real competitive advantage.
But the thesis should not be romanticized.
Not all management is bureaucracy. Good managers do more than move information. They resolve conflict, build judgment, protect culture, sequence work under uncertainty, and make calls when signals are incomplete or contradictory. AI can support those tasks, but it does not remove their human weight. A startup that eliminates management without building clarity, trust, and decision discipline will not become more intelligent. It will become more chaotic.
That is the real test.
The winners in the AI era will not be the companies that simply cut layers. They will be the ones that redesign work around clearer accountability, richer operating visibility, and stronger human judgment. AI can flatten information flow. It cannot substitute for ethics, trust, conviction, or taste.
So the future is probably not a manager-free startup. It is a startup with fewer people doing coordination for coordination’s sake, more people directly accountable for outcomes, and a much smaller gap between what the company knows and what it does.
That is the deepest implication of Dorsey’s argument. AI should not just help startups work faster inside old structures. It should force founders to ask whether those structures still deserve to exist at all.
The most important takeaway is simple: the org chart of the AI era should be built around contribution, ownership, and development, while machines increasingly handle routing, synthesis, and visibility. Startups that understand this early may not just become more efficient. They may become structurally better at learning, deciding, and adapting.