Horizontal leadership, seen from the manager’s chair

In 2002, Larry Page and Sergey Brin removed every engineering manager at Google. The experiment lasted a few months. The two founders ended up handling expense reports, interpersonal conflicts and career questions themselves, and the managers came back. Six years later Google launched Project Oxygen to establish, with data, what those positions actually contributed.

The story circulates as proof that hierarchy is inescapable. It proves something else. A manager's work exists independently of the position that carries it, and it survives the removal of the label. Horizontal leadership means writing that work down somewhere else, on named roles held by several people. Without that writing, the work travels back to the one person left, exactly as it did at Google in 2002.

This article takes the point of view of whoever has the most to lose in the operation: the manager.

Three concepts behind one word

"Horizontal leadership" is a practitioner phrase. Research offers three older and sharper ones, which get lumped together under it.

Servant leadership comes from Robert K. Greenleaf, who published the essay The Servant as Leader in 1970, after forty years at AT&T and the founding of his Center for Applied Ethics in 1964. His founding line is short: the servant-leader serves first, and the choice to lead comes after. Greenleaf says he took the idea from a Hermann Hesse novel, Journey to the East. He describes a stance, with no process and no mechanism attached.

Shared leadership has its reference definition in Craig Pearce and Jay Conger's edited volume Shared Leadership, published in 2003. They describe a dynamic influence process among the members of a group, lateral as much as downward, where several people exert influence instead of one person placed above the others.

Distributed leadership comes out of education research. Peter Gronn published "Distributed leadership as a unit of analysis" in The Leadership Quarterly in 2002, and James Spillane turned it into a book in 2006. Their contribution is methodological: they move the unit of analysis from the person to the practice, stretched over several individuals, the situation, and the tools that carry it, including agendas and evaluation protocols.

The three say different things. The first describes the intention of one leader. The second describes a team where influence circulates in both directions. The third observes a practice spread across people and instruments. A corporate speech that blends them usually promises the first while leaving the structure untouched.

What the research measures

The reference meta-analysis is the one by Danni Wang, David Waldman and Zhen Zhang, published in the Journal of Applied Psychology in 2014. It pools 42 independent samples. Shared leadership taken as a whole correlates at .35 with team effectiveness. When what gets shared is limited to the classic forms of leadership, initiating structure and consideration, the correlation drops to .18.

Two qualifications matter more than the number. The effect is stronger when the team's work is complex, which narrows the ground: a standardised production line looks nothing like a design team. And the relationship shows up more clearly on attitudes, processes and emergent team states than on hard performance. Of course a correlation leaves the direction of causality open, and a team that is doing well probably shares its leadership more easily. A solid result remains: where the work is complex, spreading influence goes together with teams that function better.

On servant leadership, the meta-analysis by Julia Hoch, William Bommer, James Dulebohn and Dongyuan Wu, published in the Journal of Management in 2018, compares the recent forms of positive leadership against transformational leadership. Authentic leadership and ethical leadership add close to nothing. Servant leadership explains variance of its own across a wide range of outcomes. It is the only one of the three that stands on its own.

What the manager's position carried, and what replaces it

What the position carried	What replaces it
Authority over everything below it	A written domain per role, decided alone by whoever holds it
Arbitration of every disagreement between teams	One named decision rule per kind of question
Filtering of information between floors	An org chart and minutes anyone can read
Promotion as the only reward	Wider roles, at the same level
"Because I say so" as the last resort	A reasoned objection, admissible from anyone

The manager loses three real things, and has good reason to defend them. They lose the right to arbitrate by default, which made up most of their perceived usefulness. They lose the information monopoly that made their presence necessary in every loop. They lose promotion as a readable path, since the rungs go away along with the levels.

They get a lot of time back. Approving leave, expenses and reports used to fill the week, and that work goes to the people concerned. They get a domain of their own, where they decide alone without asking, which a manager squeezed between two floors rarely has. And they get the right to object to a decision made by their own boss, which a management line denies them.

A team lead's Monday morning

Take someone responsible for seven people, in an organization that distributed its authority a year ago. Their inbox holds no approval requests. They hold four written roles: recruiting for their area, the relationship with two client accounts, facilitation of their team's meetings, and a representative role that carries the team's voice one level up.

At nine, they sign a 12,000 euro quote. The amount falls inside the domain of their commercial role, they decide alone, they record the decision in the team's thread and nobody signs it off. At ten, the operational meeting runs through their checklist and indicators, then handles six tensions raised by the team, two of which become tasks attached to a named role. At two, a developer wants to change the bug tracker, which touches three roles: the question goes to a consent decision at the next meeting instead of travelling up to them. At five, they notice that nobody has held the relationship with the hosting provider for two months, and they put the creation of a role on the agenda of the next governance meeting.

They decided more things alone than the year before, and they arbitrated far fewer things on behalf of others. The shift fits inside that one day.

"It's just the boss pretending"

This is the best objection on the subject, and it is often correct.

Patrick Gilbert, Ann-Charlotte Teglborg and Nathalie Raulet-Croset ran an inside study at FAVI, Poult and Chronoflex, and published their findings in Gérer & Comprendre in March 2017. They point to a paradox: flattening the hierarchy strengthens the image of the boss instead of erasing it. Once the intermediate layers disappear, every arbitration travels back to the single figure left standing. Isaac Getz's "liberating leader" is one person whose stance carries the model, and a model carried by one person ends when they do. Jean-François Zobrist left the top job at FAVI in 2009, after twenty-six years.

The article on the liberated company covers this mechanism through the French research, and the one on the problems a heavy hierarchy creates lists what a horizontal veneer leaves untouched.

Middle managers decide the outcome

Thierry Weil and Anne-Sophie Dubey studied ten organizations of 50 to 1,300 employees for La Fabrique de l'industrie, and published Au-delà de l'entreprise libérée in 2020. They describe a group left "destabilised": middle managers are asked to give up their role without being offered another one, and they rarely get trained for the one that follows. They are the people who make the transformation work or fail, and they are the people everyone forgets.

The context makes it worse. In the US, middle managers made up 29% of 2024 layoffs according to Live Data Technologies, against roughly 20% a year between 2018 and 2022. A Gusto study published in 2025 on American small and mid-sized businesses shows the number of reports per people manager doubling between 2019 and 2024, from around three to around six. A manager who hears "horizontal leadership" has good reason to understand "we are cutting my position and spreading my workload". They are often right, and they defend the old setup, which is rational.

The answer that works is to write their roles before removing their position. A manager who reads, in black and white, the four roles they will hold, with the domain and accountabilities of each, is discussing something concrete. A manager who gets announced a new stance with no document is discussing a threat.

What has to be written for authority to hold

An organization that distributes authority durably writes three things, in this order.

Roles first. Each role carries a purpose, a domain its holder decides on alone, and accountabilities the others expect. One person holds several, and a role changes when the need changes, which is what separates it from a job description.

The decision rule next. Who settles what, over which scope, and through which process when the question spans several roles. Consent decision-making asks for the absence of a reasoned objection rather than everyone's enthusiastic agreement, which keeps cross-cutting questions at the level where they arise.

The space where the structure evolves, last. A regular governance meeting creates a role, amends another, handles a tension about how the team works. Without it the org chart freezes within months and the gap with the real work widens.

The guide to transitioning to horizontal management lays out the full sequence, the article on leadership in Holacracy covers the case of a formalised framework, and the horizontal versus vertical comparison places the two models against each other.

Where Rolebase comes in

A tool creates no practice. It keeps the structure out of a spreadsheet that three people maintain by hand.

In Rolebase, everything is a role, and a role that contains others becomes a team. Every role carries its purpose, domain, accountabilities, checklist and indicators, the last two feeding the matching steps of your meetings. The "Represents its parent role" option models a team's representative to the level above. Base roles provide shared, reusable definitions such as Leader or Facilitator.

Three governance modes set who changes what. In Free mode, any member edits the org chart. In Agile mode, the person in charge of each role drives its scope and assigns members to it. In Strict mode, every structural change goes through a proposal voted by consent, unanimity or majority, and an accepted proposal applies the changes it carried. Every edit lands in a history you can undo, which makes the question of who decided what, and when, checkable. The product is open source, free up to five active members, then 5 € per user per month.

The worker cooperative EVEA grew from 60 to 140 people between 2020 and 2023, across three sites and more than eight competency hubs. Shared governance was already in place. What was missing was a map anyone could read in real time, once a hand-maintained document stopped keeping up.

Frequently asked questions

What is horizontal leadership?

Horizontal leadership describes a setup where influence and decision power spread across several people instead of following a management line. Each role carries a written domain its holder decides on alone, and questions that span several roles are handled by a decision rule everyone knows. The phrase belongs to practitioners, while research uses shared leadership or distributed leadership.

What is the difference between horizontal, shared, distributed and servant leadership?

Servant leadership, formulated by Robert K. Greenleaf in 1970, describes an individual stance where the wish to serve comes before the choice to lead. Shared leadership, defined by Craig Pearce and Jay Conger in 2003, describes a lateral influence process among the members of a team. Distributed leadership, formulated by Peter Gronn in 2002 and then by James Spillane, moves the unit of analysis from the person to the practice stretched over several individuals and the tools they use. Horizontal leadership covers all three in everyday use, with the risk of promising a stance while leaving the structure intact.

Does horizontal leadership actually improve performance?

The meta-analysis by Danni Wang, David Waldman and Zhen Zhang, published in 2014 in the Journal of Applied Psychology, pools 42 samples and finds a .35 correlation between shared leadership and team effectiveness. Two qualifications matter. The effect grows with the complexity of the work, which makes it more relevant to design teams than to a standardised line. And it lands more on team attitudes and processes than on measured performance.

What happens to a manager under horizontal leadership?

They hold several written roles instead of one position, most often around expertise, facilitation, client relationships or representing their team. They stop approving leave, expenses and decisions that belong to other people, and they gain a domain where they decide alone without asking. At Buurtzorg, what stands in for management is a set of regional coaches who each follow forty to fifty teams with no decision power.

How do you spot horizontal leadership that is only for show?

Three signs are enough. No written decision has ever gone against the boss's preference. Employees talk about autonomy without being able to quote the rule that grants it. And the setup would stop if the boss left. Patrick Gilbert, Ann-Charlotte Teglborg and Nathalie Raulet-Croset described the paradox behind this in 2017: once the intermediate layers disappear, every arbitration travels back to the single figure left, and flattening strengthens the image of the boss.

Does horizontal leadership suit every organization?

It performs best on complex, low-repetition work, in units of 5 to 40 people according to the La Fabrique de l'industrie study. Most horizontal organizations keep a conventional decision line over safety, legal, payroll and client commitments. Saying plainly what stays closed is what makes everything you open next to it credible.

The word and the thing

Horizontal leadership can be checked in one place: the list of decisions a person can make alone, written somewhere their colleagues read it. As long as that list is missing, the organization has changed its vocabulary and kept its management line, and its teams know it before its executives do.

Start with one team, and write its manager's roles before anyone else's, since they are the one who decides how this ends. See what Rolebase does, or create your organization free up to five active members.