OpenAI CEO Sam Altman states that most companies' organizational structures are derived from the military, employing hierarchical command and control structures
OpenAI CEO Sam Altman stated that most companies' organizational structures are derived from the military, employing hierarchical command and control structures.
He pointed out that observing a large number of AI agents collaborating to solve problems without leadership, competing for credit, or engaging in disputes has prompted him to rethink this model. Altman acknowledged that humans have "beautiful flaws" and will never work this way, but the teams within OpenAI that embrace AI have become smaller and changed their practices.
Altman has previously stated that successful companies in the future will heavily utilize AI, with people managing teams of agents to complete complex tasks, and he anticipates the emergence of billion-dollar companies operated by two or three people plus AI. He has noted that agents have begun to work like junior employees, with humans responsible for allocation, quality checks, and result integration.
Public reports indicate that Altman believes small teams can accomplish more work in the AI era because coordination costs have significantly decreased. OpenAI itself is also promoting a more streamlined structure and launching related agent products and platforms.
Market Mechanism: Agent collaboration reduces coordination and communication costs, enabling smaller teams to handle more complex tasks; traditional organizations relying on large-scale hierarchical management face higher operational expenses, while teams that can effectively deploy agents benefit from lower labor costs and higher execution speed.
Source: Public information
ABAB AI Insight
Sam Altman has long advocated for OpenAI's transition from a research lab to a product and agent platform, launching tools such as ChatGPT Agent and Codex, and emphasizing that people will manage teams of agents rather than simply using chat interfaces. He has repeatedly predicted that AI will enable small teams or even individuals to operate high-value companies, expressing a more pessimistic view of human judgment under pressure.
On the capital front, OpenAI is investing resources into agent infrastructure and tools, allowing external companies to complete tasks through managing agents rather than relying on large-scale internal workforce expansion. The shrinking size of teams embracing AI internally reflects that more execution is handled by agents, with humans shifting towards supervision and integration, thereby reallocating funds from personnel salaries to computing and model capabilities.
This is similar to how early software companies reduced collaboration costs through tools, enabling small teams to complete large projects, and is akin to quantitative trading firms replacing some human labor with systems. Current corporate organization is transitioning from "hierarchical command control" to "humans managing clusters of agents," with Altman positioning OpenAI to provide this foundational infrastructure.
Essentially, this represents technological substitution and industrial chain restructuring: when agents can collaborate to complete tasks without traditional hierarchies, the coordination costs of military-style organizations become a disadvantage. The mechanism lies in the fact that human collaboration involves credit competition and communication friction, whereas agents can execute in parallel according to objectives and integrate results, leading to diminishing marginal efficiency in structures that rely on large intermediate management layers, pushing companies towards models with fewer human resources and higher agent density.
ABAB News · Law of Cognition
Agents do not compete for credit, so humans need hierarchy.
As coordination costs decrease, small teams can accomplish big tasks.
Humans have flaws, which is why organizations need commands.