Every organization flattening its structure for artificial intelligence and agentic systems is running the same experiment.
Microsoft reframes leadership as hands-on proximity, with senior executives returning to direct technical contribution [1]. Meta turns managers into “org leads”. Block introduces “player-coaches”. Both redefine the role as operational partnership rather than positional authority [2]. Cloudflare and a growing list of technology companies follow the same playbook. The headlines suggest a simple story: fewer managers, lower costs, faster decisions. The subtext is more seductive: artificial intelligence will handle the coordination, and the people who once managed will move on to higher-value work.
The reality is more complex.
Eighty-four percent of organizations have not redesigned jobs to fit agentic systems [3]. They are running the experiment without changing the architecture.
The experiment is straightforward: remove layers and watch what happens. No redesign, no new decision architecture, no clarity on who decides what. Just fewer managers in the same hierarchy.
The headlines meanwhile focus on the people: who stays, who goes, who becomes a player-coach. The more consequential question is about the architecture. What decision rights replace the approval chains? What information flows replace the management reports? What accountability mechanisms replace the performance reviews?
Fewer layers in the same hierarchy always create more confusion rather than more clarity. The remaining managers ask what happens to their role. The leaders ask what happens to their structure. Both questions matter, and neither has a simple answer.
The uncertainty runs deeper than job security.
Management redistributes — from center to edge, from title to capability. The question is not whether this redistribution happens. Agentic systems make it inevitable, and your architecture can either channel the redistribution or compound the strain.
The question is what structure replaces the pyramid.
Organizations still operate on architectures built for concentrated management. Agentic systems redistribute management from a hierarchical title to a universal organizational capability practiced by every employee. The activity disperses and multiplies.
Whether this occurs is determined by structural design, not technology adoption alone [4].
The following sections examine what that means — and why structural redesign determines whether your organization channels the redistribution or compounds the strain.
Authority Moves to the Edge
In traditional structures, authority accumulates at the center. Information flows inward, decisions flow outward, and the people closest to the work wait for permission.
This design made sense in stable environments where the center could know enough to decide. It collapses when conditions change faster than information can travel.
Authority should reside where information and context exist. In adaptive organizations, that place is the edge.
Frontline workers encounter exceptions first. They hold the contextual knowledge that makes good decisions possible. Corporate Rebels have documented this pattern across hundreds of organizations: employees on the frontline often have a better understanding of products, processes, machines, and customers than people at the center [13].
Authority follows information, and information lives at the edge.
When authority is shared, decisions move faster, bottlenecks are avoided, and teams can respond more quickly to new information and changing conditions [13]. Distributed authority improves speed and responsiveness not by eliminating coordination but by relocating it to where context is richest.
Agentic systems accelerate this migration. When artificial intelligence employees perform volume work, human attention shifts to judgment, alignment, and boundary-setting. The technology trajectory itself reinforces the shift. Edge computing minimizes latency, reduces response times, and promotes real-time decision-making by processing data locally at the point of action [6]. ZEDEDA predicts that edge artificial intelligence will push decision-making autonomy even closer to the frontline [7].
The people closest to the operational reality become the natural decision-makers. Authority follows information, and information lives at the edge.
How Distributed Authority Works
Tactical autonomy is distributed authority in action. It reveals that autonomous action is infrastructure, not inspiration — organizations must build the structural prerequisites (clear boundaries, stated intent, decision architecture) before people can act autonomously [8].
Bungay frames this as the autonomy paradox: maximum freedom requires clear constraints. Clear boundaries create the space for initiative [8].
The Advice Process is a well-established mechanism for distributed authority. Anyone can take initiative and make any decision after seeking advice from those affected and those with expertise. The decision-maker listens but retains full authority. This replaces approval chains with informed judgment [11].
Rather than waiting around for information and decisions to percolate between center and edge, an organization launches thousands of initiatives and makes smart decisions every hour. Even better, they are made by the true experts on the ground [31].
You do not need to eliminate hierarchy to distribute authority. You need to redesign the decision architecture within it.
To help the true experts on the ground transitioning, David Marquet demonstrated the how on the USS Santa Fe, the worst-performing submarine in the U.S. Navy fleet. He shifted from an approval-seeking culture to intent-based leadership, training crew members to state “I intend to…” rather than asking permission. Authority migrated to where technical expertise lived. Within two years, the Santa Fe rose to top operational ranking in the fleet [5].
Marquet did not abolish functional hierarchy. He redesigned the decision architecture within it. Intent-based leadership, backbriefing, and clear boundaries replaced the permission model. The structure channeled authority to the edge without eliminating coordination.
Where Authority Already Moves to the Edge
Agentic systems make edge authority not just desirable but structurally necessary. The center cannot process fast enough. The edge cannot wait. The architecture must catch up to the technology.
This is a structural necessity already visible in practice, not a speculative future. Even highly regulated government contexts are recognizing that frontline workers hold superior contextual knowledge. The Administrative Conference of the United States (ACUS) — a federal agency that studies and recommends improvements to administrative procedure — launched a 2026 project to identify best practices for frontline decision-making in the adjudication of benefits, loans, grants, and licenses [9].
When conditions change faster than information can travel to the center and back, the only viable design is to locate decision-making where the signal originates. If government adjudication can restructure around frontline knowledge, the case for edge authority in less constrained organizational contexts becomes even stronger [9].
More (Self-)Management, Not Less
Organizations that adopt self-management, where people closest to the operational reality hold actual decision powers, discover something counterintuitive. Management activity multiplies.
Management shifts from concentrated oversight to distributed coordination. Every team member assumes responsibility for planning, evaluation, and exception handling that previously sat with a single role.
Self-management is a complete organizational operating system based on peer relationships. It replaces the assumption that one person manages many with the reality that many people manage shared commitments.
The result is more management happening, just distributed widely across the organization.
This pattern intensifies with agentic systems. Artificial intelligence handles volume work while humans move to oversight. But oversight is still management. It requires judgment, exception handling, and governance. The difference is that these activities no longer cluster at the center. They scatter to where the work actually happens.
The shift is visible in practice. Where a manager once reviewed every customer escalation, the team now handles exceptions directly using boundary guidelines designed by the former manager. Where a director once approved every campaign, the team now launches within strategic parameters and reports outcomes rather than requests.
Gartner projects that twenty percent of organizations will use artificial intelligence to eliminate more than fifty percent of middle management positions by 2026 [10].
This redistribution is an active design choice, not a passive outcome. Organizations that build support into every level make the transition sustainable.
Peer Accountability in Practice
At Morning Star, a tomato processing company, everyone is a manager because everyone is responsible for the resources needed to get the job done and for holding colleagues accountable for accomplishing the company’s mission [11]. Each person maintains dozens of direct peer relationships involving mutual accountability — far more oversight density than any single traditional manager could maintain.
Peer pressure regulates the system better than hierarchy ever could [11]. The oversight is more pervasive because it is mutual, not gentler.
Haier took this principle to scale. The Chinese appliance manufacturer transformed eighty thousand workers into more than four thousand autonomous micro-enterprises, each with full profit and loss responsibility [12]. Management did not disappear. It was redistributed into four thousand units, each practicing self-management with explicit authority and accountability.
Buurtzorg operates over ten thousand nurses in self-managing teams and achieves high patient satisfaction ratings in the Netherlands [13].
Self-management scales when the complete operating system — not just the absence of hierarchy — is in place.
These organizations demonstrate a pattern: redistribute first, then remove. Haier redesigned decision rights and profit-and-loss authority before eliminating twelve thousand middle-management positions. Morning Star built peer commitment contracts before declaring that hierarchy was unnecessary. The structural foundation preceded the structural change.
Interlocking structures, processes, and practices
Laloux makes the structural requirement explicit: self-management requires an interlocking set of structures, processes, and practices. Change only the structure, and you are left hanging in midair [11].
This is why the redistributed management model is more demanding, not less. It requires commitment contracts, conflict resolution processes, total information transparency, and peer-based performance evaluation. Each of these mechanisms replaces a function that previously sat with a single manager. The coordination does not vanish. It is rebuilt as infrastructure.
In self-managing organizations, people have to step up and confront colleagues who fail to uphold their commitments. Conflict resolution becomes the mechanism through which peers hold each other to account for their mutual commitments. This is a more direct and visible form of management, not a softer one [11].
Management activity multiplies when it shifts from concentrated oversight to distributed coordination.
As Blake Marvin observed in a comment on Fred Voccola’s LinkedIn post, accountability, decision rights, and governance do not disappear — they simply move closer to the work [14].
This is what more management looks like. More people practice oversight. More people handle exceptions. More people set boundaries. The reduction is in titles, not activity. The increase is in capability, not headcount.
From Control to Clarity
The management role transforms. Priorities for former managers can shift from operational support to strategic value [15].
As agentic systems handle routine tasks, leadership evolves toward model oversight, business integration, and workflow design — establishing the conditions within which others make good decisions rather than making the decisions themselves.
Yet strategic clarity remains the primary performance constraint. The transition from control to clarity only works when direction is explicit enough to act on. Most organizations have not built that clarity — so distributed authority produces distributed confusion instead.
Ineffective decision-making processes rank among the top constraints for the second consecutive year [16]. Leaders do not say “I cannot decide.” They say “I do not know what I am free to decide.”
When strategic clarity is missing, distributed management becomes distributed confusion. People have decision rights but no coherent direction for using them.
They ask the right questions — “Should I proceed?” “Who decides this?” “What are we optimizing for?” — and receive no clear answer. In the absence of clarity, they default to what they know: seeking approval, waiting for direction, or making isolated decisions that conflict with adjacent teams.
The question becomes whether the organization has built the clarity that makes distributed authority functional — not simply whether to distribute authority.
Team Trek frames the leader’s responsibility directly: the main job is to provide clarity as to where the organization is going, then influence the head, hands, and hearts of others to follow [17].
Gartner describes the managerial shift as moving from execution to judgment, from supervision to coaching, and from logic to empathy [10]. These are not minor adjustments. They are fundamentally different modes of contribution that require different evaluation criteria, different development paths, and different organizational support.
The Cost of Missing Clarity
Deloitte’s research on organizational decision-making shows what happens when clarity is absent. Without clear decision rights, teams experience ambiguity about who decides and confusion about accountability. Decisions are delayed while people figure out who should decide, and some decisions fall through the cracks entirely [18].
The consequence is a structural vacuum where the old approval chains have been removed and the new decision architecture has not been installed, not merely slower execution. People in this vacuum do not lack motivation. They lack the information and boundary clarity that would enable them to act. The result is paralysis disguised as caution.
If clarity is hard to maintain after intentional restructuring, it is nearly impossible to achieve by accident after layer removal. LSA Global adds the coordination dimension: lack of coordination and trust across functions is a key reason for strategy execution failure [19]. Misaligned priorities and interdepartmental conflicts create delays, duplicate efforts, and weaken accountability.
Clarity before structure
Strategic clarity concerns two things: what the organization aims to achieve and how different parts of the organization know what they are free to decide, what they must consult on, and what they should escalate. Without this cross-functional clarity, distributed authority produces siloed autonomy rather than aligned action.
Hudson’s analysis of 922 U.S. firms shows that decentralized structures perform worse than hierarchies in ambiguity – when ‘problems are not resolved via more information and deliberation, but through coalescence on a dominant narrative’ [20]. Del Sordo and Zattoni, cited by Hudson, warn that if decision-making authority remains centralized, the potential benefits of diverse team composition remain unrealized. The implication: diagnose the knowledge problem before designing the team. Strategic clarity comes first. Structure follows. The sequence matters.
The absence of strategic clarity is the structural void that makes de-layering fail. When organizations remove layers without distributing clarity, they do not reduce management. They create a vacuum where decision rights, accountability, and information flows collapse into confusion.
The Real Risk Is Structural, Not Personal
This is a structural shift, not a personality change.
Organizations that build explicit boundaries, decision rights, and provide strategic clarity enable management without managers. Those that skip this step find that removing titles simply produces confusion.
Thirty-seven percent of employees whose companies reduced management layers reported feeling directionless [21].
Structure determines whether management disappears or disperses — not technology. Organizations best positioned for agentic transformation decentralize authority, make decision rights explicit, and add strategic clarity before they adopt new tools.
The scale of de-layering is massive and accelerating. Most organizations are conducting this experiment without the structural redesign that would make it succeed.
Most failed flattening efforts treat structural change as a one-time headcount cut rather than a systemic redesign. Flat structures move complexity from managers into the system itself, requiring deliberate redesign of decision rights, information flows, and accountability mechanisms [22].
The majority of organizations still operate with jobs designed for a different era. They install intelligent tools on top of rigid hierarchies, and the results fall short because the architecture determines what the organization can and cannot do.
The evidence is accumulating. Zappos adopted holacracy in 2014 to accelerate decision-making that had slowed through multiple layers of management. The transition produced significant employee resistance and an 18 percent workforce departure — yet the company simultaneously reported a 78 percent increase in operating profit. By 2016, Zappos had begun adapting the system into a more flexible hybrid, and by 2018 had evolved toward a market-based ecosystem where teams operated as mini-businesses. As of 2025, Zappos continues to use elements of holacratic practice within a blended structure that includes team leads and clear accountability.
GitHub initially pursued a flatter model with minimal management layers, believing this would increase autonomy. As the organization scaled, coordination broke down, decisions stalled, and accountability gaps emerged. GitHub eventually reintroduced management layers after the flat model failed at scale.
The experiments demonstrates that self-management systems require customization, and that the path from hierarchy to distributed authority runs through structural adaptation, not personal abandonment.
People dynamics stay the same
Flatness does not eliminate power dynamics. In the absence of formal authority, informal hierarchies emerge based on personality, tenure, or access to leadership. Formal layer removal without structural redesign simply hides the hierarchy.
Flat structures do not eliminate power. They shift it into invisible channels where it is harder to identify, discuss, or redesign. The result is opacity, not equity.
Donella Meadows observed that changing people or parts without changing relationships produces the same behavior from different people [30]. Firing someone and hiring a replacement without changing the role, incentives, or context yields identical results. The same principle applies to removing managers without redesigning the structure they inhabited.
Laloux makes the structural warning explicit: putting self-managing structures in place without shifting mindsets creates structural incoherence — the structure and the culture must evolve together, or the organization is left hanging in midair [11].
The manager-to-employee ratio in U.S. firms expanded from roughly 1 to 5 in 2017 to about 1 to 15 by 2023 [23]. Removing just two layers from an organization with a 1 to 5 ratio could theoretically create a 1 to 125 manager-to-report ratio [23]. The mathematics of span-of-control make clear that “less management” is not what results from layer removal. The same management burden spreads across more people, often without the structural support to make it workable.
This is the arithmetic of de-layering that the headlines ignore. Fewer layers with the same span-of-control assumptions do not produce less management. They produce unmanageable spans and invisible coordination overhead. More people and more culture problems.
More than losing middle managers
Patrick Endicott describes what gets lost: middle managers historically serve as translators between strategy and execution, career advocates for team members, and integrators of cross-functional work. When these functions are removed without redesign, teams lose structured touchpoints for development and alignment [23]. The functions do not disappear. They become invisible and unmanaged.
This invisibility is what makes structural failure so dangerous. When coordination functions disappear from formal roles, they do not cease to exist. They migrate into informal networks, ad hoc meetings, and personal relationships. The organization still coordinates. It just coordinates badly.
The cost of bad coordination is not visible on the balance sheet. It shows up as delayed decisions, duplicated effort, and disengaged employees who no longer know how to navigate the organization.
This is the hidden tax of de-layering without redesign.
The cost of bad coordination shows up as delayed decisions, duplicated effort, and disengaged employees
Donella Meadows identified: the persistence of a problem despite energetic intervention is often caused by the intervention itself [30]. Removing management layers without redesigning the functions they performed is a textbook example. The intervention appears to address the symptom — too many managers — while intensifying the underlying structural problem — unclear decision rights and accountability.
This is why the risk is structural. The problem is not that the wrong people were removed. The problem is that the structure they managed was not redesigned before they left.
You might miss structural value creation
The narrative that dominates suggests the pyramid simply needs fewer layers. The structure remains a pyramid — just a shorter one with the same delays between information and decision. The problem keeps coming back because the architecture remains the same.
Hierarchical pyramids are structurally obsolete in volatile environments. They assume the center can know faster than the edge can act. Agentic systems make this assumption visible by exposing the delay between information and decision and value creation.
Organizations that cling to hierarchy while adopting intelligent tools experience this as threat. Organizations that redesign for distributed authority experience it as confirmation.
Barry O’Reilly reports that ninety-five percent of organizations are seeing zero return on investment from their generative artificial intelligence investments. Not because artificial intelligence does not work, but because their operating model does not [24].
A Carnegie Mellon simulation found that artificial intelligence agents in poorly structured workflows failed at standard office tasks – no value created. Digitizing flawed operating models does not remove dysfunction. It amplifies it [25].
High-performing artificial intelligence adopters are almost three times as likely to have significantly modified their workflows [26]. The technology produces value not when it is layered onto existing structure but when the structure is redesigned to leverage it.
The pattern is consistent across sectors and scales. Organizations that modify workflows before or alongside technology adoption capture value. Organizations that install technology on unchanged workflows capture frustration.
Siemens AG provides the positive case. The company reconfigured reporting relationships, reduced middle management, expanded frontline decision authority, and created new roles for algorithm governance and exception handling [27]. Successful artificial intelligence integration required proactive structural redesign of decision flows, not just technology adoption.
Adobe’s Kickbox program demonstrates that flexible hierarchy structures enable innovation better than rigid traditional hierarchies [28] — a pattern we see amplified when organizations integrate artificial intelligence. Flexible hierarchy is hierarchy that adapts to the work rather than forcing the work to adapt to the hierarchy. It is not the absence of hierarchy.
Siemens and Adobe demonstrate the same principle: structure determines whether technology amplifies capability or automates dysfunction.
Organizations with explicit role definitions and transparent decision rights can integrate artificial intelligence agents more naturally than those relying on informal power structures [29]. They feel value creation and confirmation.
The Control Group
The experiment — flattening the organization just by cutting middle management — has a control group. Haier redesigned authority before eliminating twelve thousand middle-management positions. Siemens restructured decision flows before expanding artificial intelligence. Morning Star built peer-based commitment contracts. Buurtzorg designed the complete operating system before scaling to ten thousand nurses. Their results are already in.
Donella Meadows described addiction as finding a quick solution to the symptom, which prevents solving the real problem [30] — adding AI to a rigid hierarchy is often exactly this. She also observed that ninety to ninety-nine percent of attention goes to parameters (settings, tools, surface adjustments), yet changing parameters rarely changes system behavior [30]. AI adoption without redesign is a parameter change: it adjusts the toolset while preserving the architecture that produces the dysfunction. The structure determines what behavior the external event releases.
The control group has already voted: Siemens restructured and saw AI augment distributed decision-making; organizations that preserved hierarchy watched technology fail. Morning Star built the complete operating system; Zappos adopted holacracy and had to iterate toward a hybrid.
This is the central lesson of the agentic era. The technology is not the constraint. The architecture is.
Build the architecture first, or automate the architecture you have. The choice is yours. The results are already visible in the control group.
Sources
- Dan DeFrancesco (Business Insider): The big strategy themes behind Microsoft’s executive overhaul — 2026
- Sarah E. Needleman (Business Insider): Goodbye, middle managers. Hello, ‘player-coaches’ and ‘org leads.’ — 2026
- Staff/Deloitte: State of AI in the Enterprise 2026 — 2026
- Tiffany McDowell, PhD & David Mallon (Deloitte): Getting decision rights right — 2023
- L. David Marquet: Turn the Ship Around! — 2012/2013
- Upadhyay, Nassar, Sandesara, Taware (CIO.com): Edge Computing: a powerful enabler for industrial frontline workers / Anna Vainer (Flolive): Edge Computing in 2026 — 2026
- Rashida Syed (ZEDEDA): 2026 Predictions: How Edge AI Is Reshaping Industrial Operations — 2026
- Bungay: The Art of Action — 2011
- Administrative Conference of the United States (ACUS): Frontline Decision Making in the Adjudication of Applications — 2026
- Kara Dennison (Forbes): Why Companies Cutting Middle Managers To Fund AI Is A Mistake — 2026
- Laloux: Reinventing Organizations, 2014; via Rod Collins (Enlivening Edge): Reinventing Management, Part 1: What Color Is Your Organization? — 2014
- GLOBIS Corporate Solutions: Rethinking the Org Chart: How Teal Models Unlock Agility and Trust — 2025
- Pim de Morree & Joost Minnaar: Corporate Rebels: The Rebel Rules for a Happier, More Productive Workplace (LID Publishing, 2020) — Buurtzorg chapter: 15,000 nurses, 1,000+ self-managing teams, high satisfaction
- Fred Voccola: AI Replaces Traditional Hierarchies in the Workplace — 2025
- Patel, Herndon, Phaneuf, Woods (PwC): No more pyramids: Rethinking your workforce for the agentic AI era — 2026
- LHH (Adecco Group): LHH 2026 C-Suite Research: Executive Turnover Falls Sharply as AI Accountability and Decision-Making Gaps Define Leadership Agenda — 2026
- W. Gary Gore (Team Trek): Organizational Clarity — 2025
- Tiffany McDowell, PhD (Deloitte): Getting decision rights right — 2023
- Tristam Brown (LSA Global): Lack of Strategic Clarity: Only One Reason Your Strategy is Failing — 2025
- Kerry Hudson, citing Del Sordo & Zattoni: Diversity and decentralization: Means or ends? Performance implications for firms and stakeholders in a new era of ambiguity — 2025
- Michelle Gouldsberry (Betterworks / Korn Ferry): The Great Flattening: What Happens When Middle Management Disappears? — 2025
- Pim de Morree (Corporate Rebels): 9 Challenges of Flat Organizational Structures (and why they are still worth tackling) — 2026
- Patrick Endicott (SAM C.E.N.T.S.): Flattening Corporate Layers Is Changing How Work Gets Led / Andy Budd (LinkedIn): Removing Management Layers: Closer to the Work or Out of Touch? — 2025
- Trisha Famor (Barry O’Reilly): Leading AI Adoption in 2026 — 2025
- Consultancy Middle East: Why enterprise-wide AI impact demands enterprise redesign — 2025
- AlignOrg, citing McKinsey & MIT: AI Success Requires Intentional Redesign of Workflows — 2026
- Andre Ripla: How AI Is Reshaping Corporate Decision Hierarchies: Organizational Transformation in the Age of Artificial Intelligence — 2026
- Baumann, Oliver; Wu, Brian (Springer Nature, Journal of Organization Design): Managerial hierarchy in AI-driven organizations — 2023
- Consultancy Middle East (Agile Dynamics): Why enterprise-wide AI impact demands enterprise redesign — 2025
- Meadows: Thinking in Systems — 2008
- Sebastian Klein: The Loop Approach: How to Transform Your Organization (Campus Verlag, 2023)
Please note: 51&even is an AI-first organization. We embrace AI at every step of our value creation and build our processes with a deep integration of human-AI capability. Humans always have the last decision. But this text was heavily built with AI.