The UNU Global AI Network Insight Blog Series is launched to celebrate the Network’s two-year anniversary. In this series, members of the UNU Global AI Network share diverse insights at the intersection of artificial intelligence and sustainable development. Bringing together perspectives across regions and disciplines, the series explores both the opportunities and challenges of shaping AI for all.
The views expressed in published articles of this series are solely those of the author(s) and do not necessarily reflect the positions of the United Nations University, UNU Macau, the United Nations, or affiliated entities. Publication does not imply endorsement or institutional approval.
Authors are responsible for the accuracy and originality of their contributions. UNU reserves the right to edit, decline, or remove submissions in line with the purpose and values of the series.
For inquiries, please contact: AINetwork@unu.edu
What happens when an organization no longer needs people to pass information upward and instructions downward? What happens when artificial intelligence can see, summarize, compare, and surface organizational reality faster than a management chain? And if hierarchy was built to solve a human constraint, what should remain when that constraint begins to disappear?
Here is a closer look at these questions.
A Corporate Experiment in Real Time
A recent experiment by a large digital-payments company brought this issue into public view. The company reduced its workforce by nearly 40%. It abolished all formal titles: Vice Presidents, Directors, Managers, all gone. It compressed its management layers to two or three, with a stated ultimate ambition of zero. And crucially, it named the replacement: not a new organizational chart, but an AI system — a real-time intelligence layer that would do what managers had always done: aggregate information, surface truth, and enable decisions, but without the distortion, politics, or career incentives that make human relay stations unreliable (Shapero, 2026; Botha & Dorsey, 2026).
In a subsequent public statement, the company’s chief executive put the logic starkly: “The most important thing we can do is get rid of the noise between the people doing the work and the people making decisions about it. AI can be that connection. Humans were never good at it anyway.”
This is not a minor organizational adjustment. It is a claim about the nature of management itself.
The Constraint That Created Hierarchy
The pyramidal hierarchy did not emerge only from a desire for control, although it often enabled control. It emerged from a hard cognitive problem: human beings can coordinate only a limited number of direct relationships before information becomes unmanageable. Organizational theorists have long described this as the “span of control.” Beyond a certain point, coherence breaks down.
The solution that every large human organization adopted was the same: add layers. But each layer compresses information flowing upward and distributes instructions downward. The pyramid is, in this sense, a lossily compressed communication network built from human neurons.
That compression solved one problem while creating another. With every relay, the signal degrades. Managers shade information to protect themselves. By the time a report reaches the top, it often bears only a family resemblance to ground truth, and the CEO is working from a picture of the company that has been sanitized, simplified, and politically processed at every stage.
Digitalization Dissolves the Premise
The premise of hierarchy was that there was no other way to aggregate organizational information at scale. That premise dissolved quietly over the past two decades and is now weakening rapidly.
Today, employees generate continuous digital traces: messages, code commits, document revisions, calendar records, and patterns of collaboration. Feeding this data stream into large language models produces something genuinely new: a real-time organizational world model, more current than a quarterly report and less dependent on human intermediaries (Botha & Dorsey, 2026).
If the information-relay function that once justified middle management has been automated, then the pipes are no longer needed.
The New Architecture: Circles, Not Pyramids
In this redesigned structure, the organizational center of gravity is not a person at the apex of a pyramid. It is an AI intelligence system that holds the continuously updated factual state of the company. Everyone works in relation to this system. Information no longer flows through ranks but is available simultaneously to those who need it.
Within this structure, people occupy one of three roles defined entirely by their relationship to outputs rather than to other people:
- Builders: those who make things. Their core competency is judgment of what is worth building. Taste, in other words, as a professional skill.
- Directly Responsible Individuals: those who own outcomes, specifically the outcomes that customers experience. They assemble the teams they need, disband them when the work is done, and carry personal accountability for the result.
- On-Field Coaches: those who are what senior leaders become when their job is no longer to review slide decks and approve budgets. They work at the front lines. Their authority is earned from demonstrated mastery.
Notably, the traditional role of “manager” no longer exists, as coordination has been transferred to the AI layer.
Learn from Earlier Attempts
The dream of manager-less organizations is not new. In 2007, Brian Robertson formalized Holacracy, a governance system that replaces hierarchical titles and reporting lines with dynamic “circles” and “roles,” distributing authority through explicit governance processes rather than designated managers (Robertson, 2015).
A well-known online retailer’s adoption of Holacracy became one of the most visible tests of this model. The results were instructive: approximately 18% of employees departed when the system was introduced company-wide, and the company eventually moderated its implementation (Feloni, 2016).
The difficulty was not ideological. It was infrastructural. Without a mechanism to automatically aggregate organizational information and maintain alignment at scale, removing the management layer simply relocated coordination costs: from managerial overhead to endless governance meetings, procedural negotiations, and the cognitive burden of self-organization.
Other organizational cases point in the same direction, though with different lessons. One European fintech company reported that its AI assistant handled a large share of customer-service interactions within its first month, performing work previously associated with hundreds of service roles (Klarna, 2024). More significantly for the organizational theory question, it restructured its remaining workforce to minimize managerial intermediation: product teams were given direct access to customer outcome data, and the management layer between teams and data was compressed to near zero.
A global music-streaming company’s widely studied “squad” model inspired many organizations seeking agility and autonomy, but later research and public restructuring decisions highlighted the persistence of coordination overhead in scaled autonomous systems (Ek, 2023; Šmite et al., 2023).
When Hierarchy Remains Necessary — And When It Becomes Toxic
A serious analysis must acknowledge that hierarchy is not always obsolete. It remains appropriate when information cannot be automatically aggregated; when tasks are highly standardized and benefit from unified command discipline; when operating environments are stable enough that predictable process outperforms adaptive self-organization; and when workforce composition does not support the high degree of individual self-management that non-hierarchical models demand.
In steel manufacturing, traditional banking operations, logistics networks, and public health bureaucracies, the pyramid remains an efficient coordination mechanism. Dismissing it wholesale would be a category error.
Hierarchy becomes counterproductive when AI can aggregate information in real time, eliminating the relay function; when value creation depends on creative judgment that hierarchical approval chains systematically delay and dilute; when competitive environments change faster than organizational structures can process; and when the scarce resource is human ingenuity rather than standardized execution. In these conditions, hierarchy does not merely add overhead. It actively destroys value by inserting distortion, delay, and political filtering between reality and decision.
The Human Contribution
This transition poses a question that every generation of technological change has asked in a new form: what is the human contribution that cannot be mechanized?
The answer emerging here is more specific than previous answers. It is three distinct capacities.
Judgment in conditions of genuine ambiguity: the determination of which rules apply, which values should be weighted, and what the right question is in the first place.
Accountability for outcomes that matter to other human beings: the willingness to stand behind a decision and bear its consequences, which requires a form of moral agency that AI systems do not possess and cannot be made to possess by technical means.
Empathy as professional competency: the ability to sense what is not being said, to read the emotional reality of a situation, to respond to another person's humanity in ways that change what is possible between them.
Govern the Transition: the SDG Dimensions
This organizational shift is not only a business story. It is a development story.
For SDG 8: Decent Work and Economic Growth, the automation of coordination functions will affect workers across the middle of organizational hierarchies: managers, coordinators, analysts, and administrative professionals. These roles have constituted much of the professional middle class. SDG 8’s commitment to full and productive employment and decent work for all cannot be realized if the structural conditions for decent work are dismantled in practice (United Nations, 2015).
For SDG 4: Quality Education, the implication is clear. Education systems that continue to prioritize rote execution and procedural compliance will prepare students for tasks that AI is rapidly absorbing. The future labor market will require critical thinking, ethical reasoning, creative problem-solving, collaboration, and the ability to work with intelligent systems.
For SDG 10: Reduced Inequalities, the risk is significant. The “judgment, creativity, and empathy” value proposition assumes a baseline of education, institutional support, and digital access that is profoundly unevenly distributed. Without deliberate intervention, this transition may amplify inequality, creating a labor market divided between a few judgment-intensive workers who benefit from AI augmentation and a larger population whose roles are eliminated without a clear path to replacement.
The Policy Agenda
These SDG connections converge on a concrete set of policy priorities.
Governments and multilateral institutions need real-time labor-market intelligence to track how AI-driven organizational restructuring is reshaping skill demand, so education and workforce systems can respond before displacement becomes structural.
International governance frameworks for AI in the workplace should establish minimum standards for transparency, worker voice, and accountability, ensuring that workers can understand the systems that evaluate and direct their work.
Development finance institutions should prioritize adaptation of educational systems in lower-income countries to the competency demands of AI-augmented labor markets. The risk that the transition to judgment-intensive work becomes a new axis of global inequality is real and quantifiable. Addressing it requires investment decisions made now.
Finally, independent research institutions such as the United Nation University should assess these organizational experiments over time, not only by efficiency metrics, but also by their effects on employment, equity, dignity, worker agency, and social trust.
Where the Work Now Is
Hierarchy is ending not because it failed, but because one of the constraints it was built to manage — the limited capacity of human beings to process information at scale — is being partially overcome. Everywhere that constraint no longer applies, the organizational forms it produced are becoming obsolete. The machine no longer needs the patch.
What the machine needs is human judgment about what it should be used for. That judgment, and the institutional frameworks that shape it, is where the work now is.
References
Botha, R., & Dorsey, J. (2026, March 31). From hierarchy to intelligence. https://sequoiacap.com/article/from-hierarchy-to-intelligence/
Ek, D. (2023, January 23). An update on January 2023 organizational changes. Spotify Newsroom. https://newsroom.spotify.com/2023-01-23/an-update-on-january-2023-organizational-changes/
Feloni, R. (2016, July 12). A former Zappos manager explains how her job changed after the company got rid of bosses. Business Insider. https://www.businessinsider.com/zappos-explains-how-her-job-radically-changed-after-switch-to-holacracy-2016-2
Klarna. (2024, February 27). Klarna AI assistant handles two-thirds of customer service chats in its first month [Press release]. Klarna Newsroom. https://www.prnewswire.com/news-releases/klarna-ai-assistant-handles-two-thirds-of-customer-service-chats-in-its-first-month-302072740.html
Robertson, B. J. (2015). Holacracy: The new management system for a rapidly changing world. Henry Holt and Company.
Shapero, J. (2026, February 28). Block laying off 40 percent of staff, citing AI advancements. The Hill. https://thehill.com/policy/technology/5758605-block-cash-app-square-parent-layoffs-ai/
Šmite, D., Moe, N. B., Floryan, M., Gonzalez-Huerta, J., Dorner, M., & Sablis, A. (2023). Decentralized decision-making and scaled autonomy at Spotify. Journal of Systems and Software, 200, 111649. https://doi.org/10.1016/j.jss.2023.111649
United Nations. (2015). Transforming our world: The 2030 agenda for sustainable development. United Nations General Assembly. https://sdgs.un.org/2030agenda
Suggested citation: Alex Lin, WorldValue New Economy, Member of the UNU Global AI Network., "AI Organizations Without Managers: When Hierarchy Becomes Obsolete," UNU Macau (blog), 2026-06-24, 2026, https://unu.edu/macau/blog-post/ai-organizations-without-managers-when-hierarchy-becomes-obsolete.
