The org chart was never really about people. AI is about to prove it.

Jack Dorsey and Roelof Botha, of Twitter/Square and Sequoia Capital renown respectively, published “From hierarchy to intelligence” at the end of last month. If you work in people, organizational design, or talent strategy, it is worth your time.

Their central argument is both historical and precise: the corporate hierarchy was not designed around how people work best. It was designed to solve an information routing problem, and it has been solving that same problem, in updated forms, for roughly two thousand years. Humans are limited in how many people they can effectively manage at once, and early organizational structures were built around that constraint. They trace a long arc to make that case: the Romans reflected it in military design, the Prussians formalized it after Napoleon’s victory at Jena in 1806, and American railroads scaled it into large commercial enterprises. Since then, most organizational innovation has worked within that same constraint, attempting to reduce friction rather than remove it altogether.

Their argument is that AI alters the constraint itself. A system that maintains a continuously updated view of an organization’s operations, priorities, and performance changes the role of management as an information routing function. That is a meaningful shift, but it also raises a set of implications that are not fully explored in their piece, particularly for those responsible for people, talent, and organizational design.

The diagnostic we don't want to skip

When organizations start collapsing management layers, and many already are, they tend to frame it as a structural decision. Wider spans of control, fewer layers, leaner org charts. The efficiency case is straightforward and the AI productivity data supports it.

Before moving on that, there is a diagnostic step worth taking seriously. What were your managers actually doing?

Span of control is a measurement of information load. Increasing it without changing the information environment breaks managers. Increasing it because the information environment has genuinely changed is a different decision with different consequences. Right now, a lot of organizations are conflating the two, moving fast on the structural change without doing the honest accounting of what they are removing.

Some of what middle managers do is coordination: status updates, alignment sessions, information synthesis, priority negotiation. AI is absorbing that work, and the case for reducing those layers is real. But some of what middle managers do is harder to replace, and far more consequential to get wrong. The job that remains is not information routing. It is thinking in more specialized and strategic ways, making the calls that require genuine expertise, contextual judgment, and an understanding of people that no system can replicate.

What remains, and what does not

Dorsey and Botha are specific about what the post-hierarchy organization looks like, and it is worth pausing on their framework before drawing conclusions. Their view is that three roles survive the transition:

• Individual contributors who are deep specialists, building and operating the core capabilities of the business. Their context comes from the system rather than a manager, which allows them to make decisions without waiting for direction.
• Directly Responsible Individuals (DRIs) who own specific outcomes with full authority to pull whatever resources they need, for a defined period, and then move on.
• Player-coaches who continue to build and do the work, while also investing in developing the people around them. Not administrators or coordinators, but practitioners who grow other practitioners.

It is a coherent model, but it is also one that emerges largely from engineering and product environments. In that context, the distinction between coordination and execution is often clearer, and the transition away from traditional management layers may be more straightforward.

What disappears in their model is the permanent middle management layer whose primary job was information routing. The system handles that. What remains is craft, judgment, ownership, and people development. It is a coherent model, but it is also one that emerges largely from engineering and product environments. In that context, the distinction between coordination and execution is often clearer, and the transition away from traditional management layers may be more straightforward. Which is exactly where the people function needs to pick up the thread.

What the model cannot fully account for

One of the more important points in their piece is the emphasis on individuals at the “edge” of the organization, those who perceive signals that are not easily modeled. This includes intuition, cultural context, trust dynamics, and judgment in ambiguous situations.

In practice, this is much of what effective managers already do when their time is not dominated by coordination. They notice early signs of disengagement before it appears in any data. They recognize shifts in team dynamics that will affect performance months later. They carry forms of institutional knowledge that are rarely documented but materially shape decisions. They operate in areas where the consequences of error are human rather than purely financial.

This work does not disappear in an AI-enabled organization. If anything, it becomes the core of the role. The risk is that organizations remove the structural layer associated with management before redefining and reinforcing these expectations, thereby losing capacity in the areas that matter most.

The implication for talent systems is direct. Performance frameworks, promotion criteria, and definitions of effective management have historically included a substantial amount of coordination work. If that work is reduced, the criteria need to evolve in advance of structural changes. Otherwise, organizations risk selecting for the wrong capabilities.

The leadership pipeline problem

There is a second implication that is less immediate but potentially more consequential. Managers are not only functional roles within an organization; they are also the primary environment in which future leaders are developed. The management layer provides a progression from individual contribution to responsibility for others, allowing for the gradual accumulation of judgment, accountability, and context. It is where many leaders first experience the complexity of balancing performance, development, and organizational needs over time. Altering or compressing that layer changes the conditions under which leadership capability is built. This is unlikely to show up in short-term metrics, but over a five to ten year horizon, it affects the depth and readiness of the leadership bench.

The venture-backed technology sector has been running an informal version of this experiment for years, promoting strong individual contributors directly into leadership roles without a traditional management apprenticeship. The results are uneven. Extending that model more broadly, particularly in larger or more complex organizations, introduces additional risk.

The question underneath the structural shift

Dorsey and Botha close with a question worth carrying into any meeting as you look ahead: what does your company understand that is genuinely hard to understand, and is that understanding getting deeper every day?

For those responsible for people and organizational design, a related question follows. What do your managers understand about your people, your culture, and your organization that cannot be modeled or replaced, and is your design strengthening that understanding or weakening the conditions that produce it?

‍

Natalia Martínez-Kalinina is an organizational psychologist and operator working at the intersection of people, culture, and business performance. Her work focuses on how organizations design team structures, leadership systems, and talent strategies to perform in moments of rapid technological and organizational change. This post was written in response to "From hierarchy to intelligence" by Jack Dorsey and Roelof Botha, published March 2026 at block.xyz.

‍