Employees, too, need both reliability and adaptability. To be effective on the job, people must have a stable working environment, access to critical resources, and clear goals and responsibilities. But they must also have leeway to adapt to changing conditions and make the right decisions in the moment—and managerial hierarchies often don’t provide that flexibility and discretion.

When you’re an executive, it isn’t easy to know the right balance of reliability and adaptability—and even if you do, it’s hard to get an organization to perform accordingly. Hence the keen interest in having organizations “feel their way” toward the desired balance through self-management, which has actually been around for decades. You could say it began about 65 years ago, when Eric Trist, an early member of the Tavistock Institute (a British nonprofit that applies social science ideas to organizational life) observed self-managed teams’ ability to substantially raise productivity in coal mines.

Back then, “longwall” mining was the unquestioned best practice. Each team performed a single task, and tasks were done sequentially—a model that fused Frederick Taylor’s scientific management and Henry Ford’s assembly lines. One team had to finish its shift before the next could start. But miners in South Yorkshire, England, began spontaneously organizing their work differently. Multiskilled autonomous groups, interchanging roles, and shifts with minimal supervision allowed them to mine coal 24 hours a day, without waiting for a previous shift to finish. In spite of that era’s prevailing belief that high productivity came with doing the same task over and over, productivity soared.

Self-managed teams took different forms as they gained popularity in the 1970s and 1980s. In Europe they became synonymous with participative management and industrial democracy. In Japan they morphed into quality circles and continuous improvement efforts. In the United States they became the organizing framework for innovation task forces. Moving to self-managed teams yielded breakthroughs in many companies, mainly in manufacturing and service operation contexts. The Volvo plant in Kalmar, Sweden, reduced defects by 90% in 1987. FedEx cut service errors by 13% in 1989. In the late 1980s and early 1990s C&S Wholesale Grocers created a warehouse of self-managed teams, which enjoyed a 60% cost advantage over competitors, and General Mills increased productivity by up to 40% in plants that adopted self-managed teams.

Such teams became more common throughout the 1990s, fueled by the promise of higher productivity in work that was increasingly complex and dynamic. In most companies that used them, just a fraction of employees were involved, generally in areas that demanded more adaptability than reliability. In time they emerged in environments where individuals could readily monitor their own performance and iteratively alter how they worked.

Eventually people wondered, Why stop at self-managed teams? After all, the heavily matrixed structures and complex reporting relationships surrounding those teams often hem them in and thwart their effectiveness. For example, when C&S CEO Rick Cohen visited Harvard Business School, more than a decade ago, to speak about his company’s success with self-managed teams, he told students that “the hardest thing is to keep the managers out of the process and just let the teams do what they do.” Why not attack the matrix head-on by applying the principles of self-management to entire institutions?

Indeed, organizations had begun to move in that direction. Management scholars Warren Bennis and Henry Mintzberg each noted a shift toward adhocracy—flexible, informal management structures—in the 1980s. A decade later the internet served as a model for what some called “the networked firm.” More recently the open-source movement, agile and scrum methodologies, and the sharing economy have inspired participative, responsive structures—holacracy, podularity (a model with roots in agile software development’s tendency to break tasks into small increments and to work with minimal planning and fast iterations), and a range of company-specific variations on self-organization. These are just the latest attempts to use self-management to reconcile reliability and adaptability.

The new forms resist hierarchical constraints—but in some ways, contrary to popular arguments, they resemble bureaucracy as sociologist Max Weber defined it in the early 1900s. Bureaucracy vested authority in depersonalized rules and roles rather than in status, class, or wealth. The idea was to liberate individuals from the dictatorial rule of whimsical bosses. Self-managing systems aim to accomplish the same thing, with less rigidity. In that sense, you could think of them as Bureaucracy 2.0.

What’s fundamentally different here? It’s how the new forms go about balancing reliability and adaptability, and the balance they seek to strike. If traditional organizations strive to be machines governed by Newtonian physics, precisely predicting and controlling the paths of individual particles, then self-managing structures are akin to biological organisms, with their rapid proliferation and evolution.

What Self-Managed Organizations Look Like

Given their origins in self-managed teams, it’s not surprising that self-managed organizations have similar codes of conduct: Members share accountability for the work, authority over how goals are met, discretion over resource use, and ownership of information and knowledge related to the work.

But what does it mean, in practice, to run a whole enterprise this way? A range of companies have made the leap, most notably Morning Star, a maker of tomato products; Valve, a developer of video games and gaming platforms; W.L. Gore, a highly diversified manufacturer; and, of course, Zappos. (For more on Morning Star, see “First, Let’s Fire All the Managers,” by Gary Hamel, HBR, December 2011.) As mentioned, we’ve seen variations on the self-organization theme, but the best-known and most fully specified of these systems is holacracy. Both because its formality makes it somewhat easier to pin down and examine and because it has been implemented more often and at greater scale than other designs, we focus significantly on it in this article.

Self-organization models typically share three characteristics:

Teams are the structure.

In holacracy, they’re “circles”; in podularity, “pods”; at Valve, “cabals”; and at many companies, simply “teams.” Whatever they’re called, these basic components—not individuals, and not units, departments, or divisions—are the essential building blocks of their organizations. Within them, individual roles are collectively defined and assigned to accomplish the work. As in traditional organizations, there may be different teams for different projects, functions (finance, tech, sales), or segments (customer, product, service). But self-managing enterprises have a lot more of them—the overall organizational structure is diced much more finely. After Zappos implemented holacracy, 150 departmental units evolved into 500 circles.

The modularity allows for more plug-and-play activity across the enterprise than in a system where teams sit squarely in particular units and departments. And the teams come and go as employees perceive changes in the organization’s needs (just as task forces and project teams in traditional organizations do, but without the surrounding matrix structure, which has a way of holding ad hoc groups together even after they’re irrelevant). Some teams are more fleeting than others. As new goals, tasks, and initiatives emerge, individuals create circles or pods or cabals to tackle them. For example, St. Louis’s public television station, KETC, mobilizes temporary teams to bring community voices and stories into programming in response to major events, such as the financial crisis and recent events in Ferguson.

Teams design and govern themselves.

Although self-organization largely avoids traditional patterns of hierarchy, teams are nested within a larger structure, which they have a hand in shaping and refining. Holacratic organizations ratify a constitution—a living document outlining the rules by which circles are created, changed, and removed. So the circles don’t just manage themselves; within those guidelines, they also design and govern themselves. The constitution doesn’t say how people should do their tasks. It explains in a broad-brush way how circles should form and operate: how they should identify and assign roles, what boundaries the roles should have, and how the circles should interact.

At Morning Star, which developed its own form of self-management, employees (in consultation with relevant coworkers) write up formal agreements known internally as “colleague letters of understanding”(CLOUs). These outline responsibilities, activities, and overall goals and contain highly detailed metrics for evaluating performance. CLOUs are essentially contracts that articulate employees’ work commitments to the organization—like annual performance previews that let your colleagues know what they can count on you to accomplish. The terms are renegotiated formally every year but can be changed at any point to reflect new work requirements and individuals’ evolving skills and interests.

Leadership is contextual.

In self-managed organizations, leadership is distributed among roles, not individuals (people usually hold multiple roles, on various teams). Leadership responsibilities continually shift as the work changes and as teams create and define new roles. Technology is essential for keeping these changes straight. In a holacracy, for example, enterprise software such as GlassFrog or holaSpirit is typically used to codify the purpose, accountability, and decision rights of every circle and role, and the information is accessible to anyone in the organization. At Morning Star, CLOUs are stored on an internal server that makes each individual’s commitments visible to everybody at the company. Transparency enables cross-team integration; all the thinly differentiated roles are easier to find than they would be in a traditional organization.

When someone isn’t a good fit for a role, it’s reassigned to someone else. Of course, assigning roles is work in itself. In a holacracy, there’s a role for that, too—the “lead link,” which also assumes responsibility for connecting a circle to the larger circles that encompass it (for instance, linking social media to marketing and communications). In more loosely defined forms of self-management, such as podularity, roles are flexibly reassigned, but it is left up to the organization to figure out how.

These three characteristics add up to an organization that is responsive to the requirements of the work rather than to the directives of any powerful individual. Traditional management goes wrong when the boss gets to prescribe what must be done—or how—because of a job description, not because he or she has particular insight into what will produce the desired outcome. Self-managed organizations strip away much of this ability to prescribe, using structuring processes (rather than a fixed structure) to maintain order and clarity.

What They Try to Accomplish on the Ground

Recent experiments with self-managed organizations have zeroed in on a few ways of improving performance. In each area, they have seen success but also problems.

Designing roles that match individual capabilities with organizational goals.

In traditional organizations, each employee works within a single, broadly defined role, and it’s often difficult for people to sculpt or switch jobs. In self-managing systems, individuals have portfolios of several very specific roles (Zappos employees now have 7.4 roles, on average), which they craft and revise to address shifting organizational and individual needs.

Negotiating with one another, employees allocate duties to those best suited to carrying them out. The process lets individuals play to their strengths and interests and serves as a safety check against roles that might be useful to one person but harmful to the team or the organization. At Morning Star, people jointly draft and adjust their CLOUs to match capabilities with work. Zappos has started a system of “badges” that let employees convey at a glance the skills they have to offer. Badges are awarded to mentees by employees already proficient as, say, a “rookie writer” (someone with limited permissions who can respond to customer-facing e-mails in times of need) or a “GlassFrog genius” (someone with a thorough understanding of the holacracy software). In a holacracy, circle members can object to a suggested role change if it would “move the circle backwards.” The person proposing the change must address the issue raised or, as a last resort, drop the proposal.

This approach to role design gives people room to grow on the job. Consider Ryan, a software developer at ARCA, a global manufacturing and services company where one of us spent more than a year observing the implementation of holacracy. Early on, Ryan—who was passionate about user interface design—saw an unmet need to ensure that ARCA’s software had a consistent look and feel. At one of his team’s structuring meetings, he pitched the idea of creating a role for this work: UI liaison. No one in the circle thought this would cause any harm, so the role was created and the lead link assigned it to Ryan, who also continued to fill the software developer role. This allowed him to simultaneously improve the group’s performance and pursue a professional growth opportunity.

Unlike fixed structures built around specialists who dedicate themselves to one function full-time, these new organizational forms let employees become “utility players,” with highly focused roles they can fill in multiple areas of the business. Take Karl, who came to ARCA before it implemented holacracy. A recent law school graduate, he had little business experience but showed great potential with his legal and analytical skills. His versatility allowed him to take on multiple roles at the growing company, in sales, legal services, and operations. However, as he worked across functional groups, he felt his contributions were getting lost in the organizational structure. When the company adopted holacracy, Karl’s many roles across multiple circles became explicit and visible. He thought his value was more clearly recognized, which gave him even more confidence to initiate changes and make decisions. Karl said, “Pre-holacracy, I felt pretty empowered but always ran stuff by people. I think an org implementing holacracy is saying, ‘You don’t have to run stuff by us anymore.’ I’ve taken the opportunity to exercise more judgment and discretion.” As one of his peers noted a few months into the new system, “Holacracy has really expanded his influence in the company.”

How did Karl fit all this work in? Holacracy let him jettison roles that weren’t a good use of his time. For instance, he used the structuring process to carve out some administrative responsibilities and pitch them as a separate role, which the lead link filled with an enthusiastic new hire. Although this shift in responsibilities was initiated by an individual contributor, not by a manager, it was highly formalized and official.

The upside of designing roles in this way is straightforward: Because employees are driving the process, they have a greater sense of making real progress on meaningful work. Teresa Amabile’s nearly 12,000 data points on the quality of “inner life at work” show that having a daily sense of forward movement—even the smallest wins—along with colleagues who provide resources, advice, and help are by far the two most significant factors differentiating good days from bad. These factors are strongly associated with creative problem solving, motivation, and engagement. Although studies of the effects of self-managed teams on employee engagement have shown mixed results, self-managed organizations are explicitly designed to remove impediments to day-to-day progress in everyone’s work and to set colleagues up to be positive “catalysts” for one another.

Assuming that the connection is borne out, is the shift from traditional jobs to a larger number of microroles a net benefit? Possibly—but role proliferation has costs, too. It creates three kinds of complexity, all related to human capital:

First, it complicates actually doing the work, because employees struggle with fragmentation. A significant body of literature on goal setting (aptly summarized by Marc Effron and Miriam Ort in their book One Page Talent Management) finds that employees perform less well on each goal as they take on more beyond just a handful. At Zappos, each of the 7.4 roles an individual fills contains an average of 3.47 distinct responsibilities, resulting in more than 25 responsibilities per employee. People grapple with where to focus their attention and how to prioritize and coordinate across circles—even with simple scheduling issues. To partially address these challenges, Zappos is trying out a tool (modeled after ordinary budgeting systems but expanded beyond dollar amounts or head-count limits) called People Points: Each circle gets a certain number of points with which to recruit individuals into roles, with senior management determining the points by assessing the business value of the circle’s work. (The company is exploring crowdfunding models to replace this top-down budgeting.) And each Zapponian gets a budget—100 points to allocate as he or she chooses. The system serves as a marketplace for the work that needs to be done, allowing a person to work across multiple teams without being told where to work. It also puts the onus on employees to fill their time with valuable roles.

Second, having so many roles complicates compensation. As people assemble their personal portfolios of roles, it becomes difficult to find clear benchmarks or market rates. For instance, what would you pay someone who divides her time between developing software, serving as the lead link for a software development team, working on marketing strategy, creating internal leadership training, doing community outreach, and planning events? Zappos is experimenting with basing compensation on the acquisition or application of its skill badges. But the complexity is still daunting.

Third, role proliferation complicates hiring, both into the organization and into particular roles. Although new employees are brought on to meet specific needs, they quickly start adding other roles to their portfolios. In the last three months of 2015, Zappos’s roughly 1,500 employees made and received 17,624 role assignments (11.7 per employee), or about 195 per day. Given that volume, the company developed Role Marketplace, a tool to quickly post open roles and manage applications, with lead links ultimately deciding who fills the roles. The tool handled almost a quarter of those 17,624 assignments. Using both People Points and Role Marketplace, an employee could potentially find, apply for, be assigned to, and start working in a role within a single day. That’s a lot of activity to keep track of, even if you’ve got software to help.

Making decisions closer to the work.

Self-management aims to reduce the red tape and endless sign-offs usually needed to make decisions in bureaucracies. In traditional organizations, intricate webs of titles, job descriptions, and reporting relationships can make it difficult to figure out who decides what. In some of the newer models, such as holacracies, everyone can see who holds each role and what people are responsible for. The processes and norms for decision making are streamlined too. Rather than run ideas up the flagpole and wait for answers to come back down, individuals go directly to the people who will be affected. Within holacracies, this is known as “going role to role.” It means that messages are less likely to get watered down or misinterpreted through layers of management.

Communication is supposed to become more efficient and accurate as a result, which is good for reliability. But to make smarter decisions in this sort of system, all members must exercise their power and voices, which doesn’t always happen. One misconception about self-managing organizations is that they eliminate differences in status. Although those differences may be mitigated, they still exist and must be managed. Some people have more power than others, and managers who used to supervise certain activities may at times try to reassert control, making it hard for employees to know whether to follow the new system or listen to their old boss.

It can also be tough for people to “step up” and claim their power. An employee at ARCA, observing that members of her circle were not challenging a top-down order from a former boss, said, “I feel that employees haven’t explored their agency within holacracy.” For such agency to thrive, both managers and subordinates must unlearn old behaviors. Another employee, who formerly had a managerial title, talked about how much time he used to spend approving others’ decisions. Since the move to holacracy, he’s had to shift to enabling mode, encouraging individuals to make decisions on their own.

Self-managing systems reinforce this unlearning to a point, through training on how to “work in” and “work on” the structure and through processes and norms that make it difficult for earlier forms of power to reemerge. For instance, 12 months into Zappos’s implementation of holacracy (and just a couple of months before the severance-package offer), 400 employees had completed the three-day holacracy training, and 90 had become “certified facilitators” of governance and working meetings. But trainings don’t in themselves eliminate problematic behavior, such as micromanaging others or infringing on the autonomy of former subordinates. Old power rules can be deeply embedded in culture and institutions and may require continual attention to unravel.

Even if employees want to speak up, it can be hard to absorb all the rules of engagement—and once people start applying them, that “structuring” work can feel almost as onerous as the Byzantine hierarchy it replaced. If every circle has a monthly governance meeting, as is common in holacracies, and if employees are in 4.1 circles, on average, the meeting time adds up. Zappos employees have so far dealt with the challenge by making their meetings more efficient and using technology to reduce the need for direct interaction. For example, the company developed a Slack bot to run governance meetings according to holacratic rules. Although the automated facilitation and virtual discussions through Slack reduce the time investment, the structuring work is still relentless, with each person involved in roughly one governance conversation a week. At Medium, the social media company that stopped using holacracy, that work proved too much to sustain. Doyle said in his blog post, “The system had begun to exert a small but persistent tax on both our effectiveness and our sense of connection to each other.”

Responding to emerging needs in the market.

We’ve traditionally romanticized our leaders as scouts with keen vision who monitor the horizon for developments that deserve the attention of the organization and its people. And thanks to increasingly advanced analytics, leaders’ observations have become far more precise. Yet a great deal ofevidence shows that efforts to drive change programmatically from the top, solely in response to what senior leaders see, often fail.

Self-managing organizations take a different approach. Consider how Valve made the decision to expand from PC games to hardware. The company’s 400-plus employees self-allocate 100% of their time to projects they feel are valuable to customers. They collaborate in cabals, which people form and re-form, project by project, by wheeling their desks together, often several times a day. When a few employees got sufficiently tired of repeated customer requests for hardware that would let people play games in the living room, they formed a cabal to investigate the idea. When others recognized Valve’s potential strategic vulnerability to a “closed” Windows store, they allocated some time to that issue. In neither case did a siren sound from a lookout on high; the problems were detected and addressed on the ground, through a steady accretion of talent. In November 2015 the cabals facilitated one of PC gaming’s largest hardware releases of the year, built on an open platform that signals Valve’s willingness and ability to protect against the threat of being closed out of customers’ PCs.

It’s possible, however, to be too responsive to your customers. Steve Jobs famously pointed out that the market doesn’t always know what it wants. AsBain’s research on growth through simplicity shows, adding SKUs in response to perceived customer needs can mean less revenue. And Bob Moesta, whose Re-Wired Group advises organizations on demand-side innovation, distinguishes between what customers explicitly ask companies to supply and a more holistic view of demand. He says the latter is where real value is created—but organizations need a level of reflection that goes beyond simple responsiveness. Although it’s important to be close to your customers, it’s also critical to maintain a broader perspective so that you don’t follow them off a cliff.

You might assume that the three goals of self-management structures—designing roles that match individual capabilities with organizational goals, making decisions closer to the work, and responding to emerging market needs—would make leaders less relevant. Yet one of the greatest challenges of implementing the goals at scale is insufficient leadership. When leadership is a shared responsibility, everyone must understand and practice it. You end up with more formal team leaders as the number of modules increases. Since adopting holacracy, Zappos has gone from 150 team leaders to 300 lead links, who are responsible for its 500 circles.

Of course, managing looks different in these structures. It’s less about supervision and direction and more about designing, facilitating, and coaching. One former manager at ARCA said, “Leadership might be even more important in a holacracy than in a traditional management structure. You have to lead by example and round up the troops rather than rely on authority.” Members of self-managing teams have been saying similar things for decades.

Costs and Benefits “Above the Trenches”

So far we’ve looked at self-management on the ground. But how does it work at a macro level—where organizations set strategic direction, oversee global operations, and shape their overall performance and trajectory?

Consider a large consumer packaged goods company like PepsiCo. Suppose it is deciding whether to shift the ingredient mix of a product made for a certain market in response to consumer demand for fewer artificial sweeteners. Self-management naturally facilitates such changes. The people who touch the decision get together, evaluate the opportunity, sort out the practical details (for example, discontinuing the use of certain suppliers), and then make it happen, all without interference from above.

However, a company of PepsiCo’s scale has a multifaceted operational agenda, which may include simplifying the global supply chain and freeing up capital for acquisitions. To achieve such goals, you’ll need more than an array of small, local moves. In fact, you’ll probably have to take actions that are suboptimal in a number of specific contexts. For example, consolidating suppliers will cut complexity and costs overall, but you’ll miss out on certain niche suppliers who could offer higher quality and lower prices in emerging markets. Structures that transmit guidance from the top are better equipped than self-managed organizations to make local trade-offs in service of scale—a critical advantage for a global CPG company.

This is doubly true when it comes to corporate strategy. Most executives view strategy as an essential feature, but proponents of self-managed organizations—like holacracy cofounder (and coder) Brian Robertson—argue that it is actually a bug. Robertson writes in his book about holacracy, “When you impose a ‘should’—as in ‘I should be X in five years’ time’—you create an attachment to that outcome; the attachment limits your ability to sense when reality is not going in that direction, or when other possible opportunities arise that might conflict with what you first set out to achieve.” Because it’s a ground rule of holacracy that you can revisit any decision whenever you want, Robertson adds, “you will find it very difficult to drive others’ behaviors on the basis of targets defined in advance.” Even though strategic planning isn’t explicitly prohibited, such plans are often replaced by continually updated rules of thumb that take the form of “emphasize X, even over Y.” At Zappos, providing the best customer service and increasing short-term profits are both guiding principles—but if employees ever have to choose between the two, they know to pick customer service.

In our view, this approach to establishing direction isn’t viable for certain kinds of organizations. Take Sirius XM: It has invested billions of dollars to create a satellite radio infrastructure that will generate returns for decades, so it needs a clear, stable, consistent overarching strategy. W.L. Gore, with its portfolio of technology innovations that can be brought to market in different ways, requires much less top-down strategic maneuvering. Zappos is probably somewhere in between. Its strategic positioning has been clearly differentiated for many years—and much of what has transpired since it adopted holacracy represents an extension of that. However, Zappos has adapted to a shifting market by making significant changes in product mix, customer targeting, and pricing. Working within the framework of holacracy, the company achieved a 75% year-on-year increase in operating profit in 2015 as a result of those strategic moves. So, although it’s hard to say whether holacracy would enable the company to navigate major changes in the competitive environment, early signs are promising.

Finding the Right Amount of Self-Management

As we’ve learned from self-managing teams, blanket arguments for or against broadly applying the principles of self-management miss an important point: Most organizations, particularly large corporations, should adopt these techniques in part, not in whole. We’d be surprised if more than 20% of the Global 1000 looked “teal” in 2030, to use Frederic Laloux’s term for “whole,” evolutionary, self-managing organizations. But we’d also be surprised if more than 20% didn’t significantly draw on some of the techniques within their corporate frameworks.

A great deal of piecemeal adoption is already happening. Procter & Gamble, for instance, operates a complex matrix organization in order to integrate its many brand categories, geographies, and functions. But it also has a vast open-innovation program, in which teams of people outside P&G’s walls organize themselves to solve problems for the company. Google and 3M provide familiar examples as well: For decades employees have been encouraged to devote a percentage of their time to self-directed work—a volunteer economy that exists alongside the managerial hierarchy’s more directed economy.

Deciding where to apply self-management in an organization hinges on three questions: What needs to be reliable? What kinds of adaptation are important? And what organizational forms will produce the right balance in this case?

Using self-management principles to design an entire organization makes sense if the optimal level of adaptability is high—that is, if the organization operates in a fast-changing environment in which the benefits of making quick adjustments far outweigh the costs, the wrong adjustments won’t be catastrophic, and the need for explicit controls isn’t significant. That’s why many start-ups are early adopters. The business of designing and developing games also fits these criteria well, as Valve discovered. But in reliability-driven industries such as retail banking and defense contracting, hierarchical structures prevail, even if there is room for niche competitors (in banking, think of Umpqua, famous for having a phone in every branch that enables customers to ring the CEO’s office) or for certain units within the organization (such as the original Skunk Works at Lockheed Martin) to go against the traditional grain.

Companies must also work out how much hierarchy and process they need to ensure coherence and what other kinds of “glue,” such as shared purpose and a common ethical compass, they can use. Dov Seidman’s “The HOW Report” quantifies the degree to which various companies rely on those other cohesive elements and links self-governance to a range of performance outcomes. Seeing how others have fared can help organizations sort out whether—and where—this particular glue makes sense for them.

Ultimately, and somewhat ironically, the next generation of self-managing teams is demanding a new generation of leaders—senior individuals with the vision to see where it is best to set aside hierarchy for another way of operating, but also with the courage to defend hierarchy where it serves the institution’s fundamental goals.