When a CEO tells me they wish their employees were more capable of “self-organizing,” I usually ask what exactly that means. The answer matters for more than academic precision; leaders drawn to the notion of decentralization routinely turn to self-organization as the mechanism that will deliver it. The appeal is clear: Give people more autonomy, reduce bureaucracy, and allow coordination to emerge naturally rather than through layers of management. In other words, let them “organize themselves.”

But this appeal rests on a confusion. As I wrote in a recent paper, decentralization and self-organization aren’t the same thing; the difference comes down to a simple question: What does the structure of influence look like?

Decentralized vs. self-organizing systems

In a decentralized organization, influence is distributed in an egalitarian manner. Decisions are made by many people rather than concentrated at the top. But behavior throughout the system is shaped by the organization’s structures and processes that affect everyone: shared norms, incentives, governance mechanisms, hiring practices, cultural expectations, operating principles, and so on. 

Influence, in this sense, is still global. Open-source software communities, Wikipedia, scientific communities, worker cooperatives, and many decentralized autonomous organizations (DAOs) sit in this family. More specifically, firms like entertainment software and technology company Valve Corp., and materials science company Gore, fall into this category. Although formal hierarchy is limited or even absent, a global structure of influence is very much present.

In a truly self-organizing system, no person, team, or overarching structure guides the whole system. Instead, patterns emerge solely from how individuals interact with one another (i.e., local interactions). Consider a flock of birds: Each bird responds only to nearby birds, with no leader and no overarching mechanism directing the flock. Yet, coordinated behavior emerges. Termite mounds, ant colonies, and unmarked “lanes” forming spontaneously among pedestrians work the same way.

It’s easy to confuse the two ideas because they often appear together in the same examples in the natural world. The difference becomes obvious only when you look closely in the social world.

Unblurring the line

Markets offer a good illustration of a highly decentralized system, with millions of participants making independent decisions. But market participants are influenced by property rights, contracts, regulations, currency, courts, and social norms—systemwide structures that shape behavior and outcomes. Therefore, markets may be decentralized, but they aren’t purely self-organizing. Managers who opt for “just let things sort themselves out” tend to forget how much policy scaffolding the market depends on.

The same principle applies inside organizations. Valve Corp. is often cited as an example of radical decentralization. Employees choose projects, form teams voluntarily, and move between initiatives with considerable freedom. Yet the company still depends on hiring practices, cultural norms, shared expectations, and organizational principles that influence behavior throughout the organization. Moreover, the founder played a role in designing this system and continues to contribute actively, particularly through hiring. 

As such, the resulting order doesn’t emerge solely from local interactions but depends on structures that shape how those interactions unfold. Unlike the “invisible hand” of pure self-organization, this is closer to a hidden hand: a global influence structure embedded in the design of the system rather than exercised continuously through overt command. The hand, though less visible than in a traditional hierarchy, is doing the same kind of work. In its absence, the result may be chaos or an undesirable order.

DAOs make the point more starkly: These are organizations built explicitly to eliminate hierarchy, with governance handled through token-weighted voting on a blockchain. They are perhaps the most ambitious attempt in recent times to engineer pure decentralization through self-organization at scale. However, large-scale evidence on thousands of DAOs shows something striking: Systems designed to eliminate hierarchy reliably reconcentrate influence, often through the accumulation of governance tokens by a few people. When you design for pure self-organization, you usually don’t get it. Instead, you get a structure shaped by global influence—in other words, a structure in a different costume.

Real lessons from biology

Why are the analogies of self-organizing flocking birds and termite colonies misleading for managers?

Flocking birds, ant colonies, and immune systems are genuinely self-organizing, and also genuinely adaptive. But this is due to a process that doesn’t transfer to human organizations: evolution through natural selection. The order in an ant colony is impressive because of how the local rules that ants follow are fine-tuned over a long time, so that maladaptive colonies die out. What we observe today, and find so fascinating, is the small surviving subset of designs that produces useful global order. The invisible hand looks competent because natural selection has hidden its failures.

Unfortunately, organizations don’t have the luxury of a patient selection process over an extended time frame. When people interact locally without coordinating structures, the resulting order may or may not be productive. Self-organization can bring innovation, adaptability, and collaboration. On the flip side, it can also lead to silos, the duplication of effort, polarization, and decision paralysis.

Economist and policy expert Thomas Schelling showed how even mild local preferences can generate segregated neighborhoods—even if that outcome is unintended by any individual. Online communities often self-organize into echo chambers rather than productive, deliberative spaces. The order is real. The adaptiveness isn’t guaranteed.

A matter of design

This is where organizational design becomes critical, and where the role of leadership in decentralized organizations is often misunderstood. Leaders don’t necessarily need to direct day-to-day decisions. Their more important role is to create the conditions under which decentralized decision-making can succeed: defining boundaries, establishing rules, shaping incentives, building culture, determining membership, and creating mechanisms that help people coordinate. 

Done well, the hand becomes hidden, not absent. This means the need for constant intervention is reduced as the system becomes capable of coordinating itself within carefully designed constraints. The challenge of organization design for decentralized systems, then, isn’t whether to eliminate systemwide global influence altogether, but to determine the extent to which influence can be distributed evenly yet remain effective.

Even the financial market, one of the most decentralized systems ever created, depends on institutions that shape behavior throughout the system. There is little reason to expect organizations to succeed through pure self-organization alone.

So, the next time someone in your company proposes a “self-organizing” solution, ask the following three questions:
• Who controls entry into the system, and what kinds of people, capabilities, and behaviors does that selection process produce?
• How does the system distinguish adaptive from maladaptive patterns once they emerge, and how quickly can it amplify the former and suppress the latter?
• What structures, norms, incentives, or actors are shaping behavior throughout the system, even if nobody refers to them as “authority” or “hierarchy”?

Published July 16, 2026, by INSEAD.