Return-to-office fights are framed as culture wars, but the real contest is over authority. Hybrid schedules, AI sprints, rolling restructurings: each exposes a simple fact. When the stakes rise and the clock runs, someone must decide. Many firms try to get the speed of networks without the weight of hierarchy. Then a launch stalls, a customer crisis hits, and an unofficial boss appears anyway, often late and unaccountable.
This is solvable. The choice is not hierarchy or network. It is whether you design the switch between them.
What networks do well, and what they do not
Self-managing organizations decentralize authority in a radical, formal, and systematic way across the whole company. Their distinguishing feature is the elimination of the manager–subordinate relationship (Lee and Edmondson, 2017). The management literature they synthesize makes a clean distinction. Managerial hierarchy performs reliably when work is known and conditions are stable. It tends to falter when teams must solve complex, nonroutine problems that span functions. In those cases, rigid layers and status differences can suppress initiative and learning, and slow cross-boundary collaboration (Lee and Edmondson, 2017).
That is the network promise: local decisions by the people closest to the work, fewer status barriers, more fluid problem solving. It is real, and it is valuable. It is also incomplete. Few firms live in the steady-state conditions where either hierarchy or network rules alone. Most need both, often in the same week.
The bosses never left
The flat ideal often conceals what actually happens in practice. Organizations that present as bossless commonly retain real bosses and hierarchies, just informal or renamed. Under dynamic and time-sensitive conditions, these firms struggle with innovation, scale, and staying power (Foss and Klein, 2022). The creative software company Valve is the cautionary tale. Employees described the firm’s 2013 state as “The Wilderness,” a period of drift under its flat model. The flagship title Half-Life: Alyx required management intervention to ship. Foss and Klein also report that Valve lacked dedicated customer support and marketing departments, which they connect to its commitment to flatness. The consequence was a system that excelled at exploration yet stalled when coordination and commitment were needed (Foss and Klein, 2022).
The lesson is not that hierarchy always wins. It is that when authority is fuzzy, it reappears informally. Decisions get made by influence, tenure, or personality instead of by a clear mandate. That slows action and muddies accountability, the opposite of what networks are meant to deliver.
Zappos moved from manifesto to mechanism
The most visible corporate experiment with holacracy shows the arc. Zappos was, in 2016, the largest company to implement the system fully, a fact that fed polarized debate about “bossless” management (Bernstein, Bunch, Canner and Lee, 2016). Then the costs surfaced. In 2015 CEO Tony Hsieh gave employees a choice to commit to holacracy or take a buyout, and 18 percent left. By 2017 the company had replaced holacracy with an internal market in which teams acted as small businesses with their own P&Ls and traded with one another at market prices. Along the way it eliminated holacracy’s meeting bureaucracy and brought managers back, retaining only a circular hierarchy as a loose artifact. The company’s leaders cited business performance challenges and criticism that holacracy was too inward-focused (Groth, 2020).
Strip away the labels and the move is pragmatic. Zappos shifted from a governance manifesto to operating mechanisms. It reinstated explicit authority and market-like incentives where the network had bogged down. That is a pattern worth copying: keep autonomy where it speeds learning, and restore clear decision rights where coordination and pacing matter.
Build a toggle, not a truce
Managers matter, not as ornament but as function. Foss and Klein distinguish two forms of managerial authority. One version, call it Mark I, includes hiring, firing, and giving direct orders. The other, Mark II, is about designing structures, setting direction, and building consensus. Their argument, summarized in strategy+business, is contingential: near-bossless firms do not show systematically superior performance, and there is no one superior structure for all contexts (strategy+business, 2022). That is a useful design cue.
Use Mark II authority to run the network most days. Make teams accountable for outcomes, let roles be flexible, and push decisions to the edge. Then specify the exact conditions that flip the system into Mark I. Examples: a sev-1 customer incident, a regulatory deadline, a board-approved bet at risk. Name the incident commander role in advance, state their temporary powers, and define the handback to normal operations. Publish who can staff up, reallocate budgets, break ties, or stop the line. Practice the switch in drills, the way safety-critical organizations rehearse emergencies.
Two more design moves follow from the evidence:
- Make authority visible, even when distributed. Lee and Edmondson’s core definition hinges on eliminating the manager–subordinate tie. If you adopt that, be explicit about what replaces it. Who can change roles, who adjudicates conflicts across teams, and where appeals go are not bureaucratic footnotes. They are the backbone that keeps a network from sliding into the shadow hierarchies Foss and Klein describe (Lee and Edmondson, 2017; Foss and Klein, 2022).
- Protect coordination functions. Valve’s gaps in customer support and marketing illustrate what happens when essential connective tissue is treated as optional because no one wants to be the “boss.” Charter these functions clearly and tie them to outcomes, even if the rest of the system is fluid (Foss and Klein, 2022).
The practical read for the 2020s: do not pick a side in an ideology war. Pick a clock speed and a clarity level for each kind of work, then wire the controls to switch between them. Networks without a failsafe drift. Hierarchies without a release valve calcify. The advantage goes to companies that can do both on purpose, and on time.
Sources
- Lee and Edmondson (2017). Self-managing organizations: Exploring the limits of less-hierarchical organizing. Defines SMOs and synthesizes when hierarchy helps or hinders, grounding the network–hierarchy tradeoff. https://www.sciencedirect.com/science/article/abs/pii/S0191308517300059
- Foss and Klein (2022). Who's the boss? The persistence of entrepreneurial hierarchy in flat organizations. Shows that flat firms retain informal hierarchy and struggle under dynamic conditions, illustrated by Valve’s drift and missing functions. https://link.springer.com/article/10.1007/s41469-022-00132-8
- Groth (2020). Zappos has quietly backed away from holacracy. Reports the 18% buyout, the rollback of holacracy’s mechanics, the return of managers, and the shift to an internal market with team P&Ls. https://qz.com/work/1776841/zappos-has-quietly-backed-away-from-holacracy
- strategy+business (2022). Can bossless management work? Summarizes Foss and Klein’s view that near-bossless firms are not systematically superior and distinguishes Mark I vs Mark II authority for contingent design. https://www.strategy-business.com/article/Can-bossless-management-work
- Bernstein, Bunch, Canner and Lee (2016). Beyond the Holacracy Hype. Notes the polarized debate and that Zappos was the largest firm to adopt holacracy at the time, framing the case’s significance. https://hbr.org/2016/07/beyond-the-holacracy-hype