The Pendulum Stops Here

Rethinking How Organizations Coordinate

by Janka Krings-Klebe and Jörg Schreiner

Most large organizations have reorganized more times than their leaders can comfortably count. Decades of restructuring show a familiar pattern: decentralize to unlock speed and innovation, then recentralize to tighten alignment and efficiency. Loosen, tighten, loosen, tighten. When performance falters, the instinct is to pull control back to the center. When growth returns, autonomy is restored. The pendulum keeps swinging because, despite decades of organizational research, the underlying downsides of centralization and decentralization have never been properly addressed. At the core lies a simple problem. The benefits of local competence and agency work best at small scale. Yet as the number of interdependent contributors grows, so does the need for tighter integration.

How can organizations preserve the local competence that makes each unit effective while achieving the integration that enables the whole enterprise to deliver value? Lawrence and Lorsch named the problem in 1967. Henry Mintzberg (Thinkers50 Lifetime Achievement Award 2015) mapped its structural consequences. Since then, each generation of management thinkers has offered a fresh answer: flatter hierarchies, cross-functional teams, agile frameworks, autonomous units. Yet the tension persists. The reason is straightforward. Most solutions try to solve the problem by standardizing operations. Organizations need something else: standardization at the interface.

That shift changes the logic of coordination. Teams no longer need to work in the same way to work well together. They need to be compatible in how they connect: how commitments are negotiated, how outputs are specified, and how information, goods, and resources move across boundaries. That is the architectural shift from hierarchy to ecosystem.

Why Culture Alone Cannot Scale

As organizations grow beyond a handful of teams, coordination costs rise sharply. In a small team, coordination feels effortless because the group quickly develops their own cultural code to facilitate communication, sharing and handover. Edgar Schein (inducted into the Thinkers50 Hall of Fame in 2009) observed that culture is essentially the accumulated learning of a group, the shared solutions to recurring problems, internalized over time as norms and assumptions. In a team of five or ten people, coordination patterns are triggered through their well-known cultural codes and communication protocols, almost subconsciously. The team operates under the same constraints, experiences are shared, feedback is immediate, and learning is collective. When a problem arises, everyone can see it. When a solution works, everyone learns about it. Culture emerges from this tight feedback loop. It works brilliantly, because coordination efforts and outcomes are tightly connected, visible and shared.

The moment a second group is added, this logic begins to fracture. Constraints diverge. Experiences diverge. Each group develops its own well-functioning cultural logic, and those logics become mutually opaque. The very success of culture at the team level becomes an obstacle to coordination between teams. This is less a management failure than a structural consequence of how learning works.

From the pyramids to moon rockets, as soon as more than one group is needed to create value, organizations require a mechanism that helps them organize contributions across group boundaries. History’s answer was hierarchy.

The Cost of the Hierarchical Solution

Hierarchy solved a specific problem efficiently, but its costs were long treated as unavoidable. It emerged as an answer to a specific constraint: how to move coordination signals across many groups when information is expensive to transmit. A hierarchy routes information upward to a point with visibility across units, a decision is made, and instructions travel back down. Although it worked, its costs were largely treated as unavoidable: latency, distortion of purpose, and the concentration of decision-making capacity at nodes increasingly remote from the actual work.

Hierarchies work best when they can create commonality at scale, so that is what they inevitably do. They reduce the variety of solutions to the coordination problem by standardizing processes, metrics, control systems and incentives. Over time, this bureaucratic effort to align diverse contributions becomes increasingly disconnected from value creation. Lawrence and Lorsch described the consequence clearly: the more an organization reduces differentiation to ease integration, the more it degrades the local competence that makes each unit’s contribution worth having. And conversely, the more it allows differentiation to build local excellence, the harder integration becomes. This is the tension they identified. It is structural. It cannot be optimized away.

Every major attempt to escape this bind has largely stayed within this same underlying logic. For example, cross-functional teams add integrative structures at the seams yet leave the hierarchical governing structures outside intact. Agile scaling frameworks try to coordinate autonomous teams, but without changing the basic standardized operation architecture. In practice, they just reintroduce coordination hierarchy by other means. Approaches that consciously avoid hierarchy, such as Buurtzorg (founder Jos de Blok was the recipient of the 2019 Thinkers50 Ideas Into Practice Award) and W.L. Gore, limit the number of teams in a value stream and therefore reduce the coordination burden. This works as long as the number of contributions that need to be aligned and integrated to create value can be limited. Yet it does not fully solve the scaling problem.

The limitations become visible when value creation requires fluid recalibration of coordination efforts at large scale. In today’s highly volatile business environment, that is the challenge that many large organizations now face. The very ground appears to be shifting, and the managerial instruments for alignment, integration and control increasingly lose their effectiveness.

Standardize Interfaces, Not Operations

Previous approaches sought coherence through common processes, metrics, roles, and management systems. The breakthrough alternative is to standardize the interfaces used for coordination, not the operations. This is a much more flexible approach, and one that has proven its worth in the domain of engineering over and over again. Consider what happened in software architecture when the industry moved from monolithic systems to micro-services. In a monolithic product, internal consistency has to be enforced and upheld through centralized controls. They ensure smooth operation, but they need to know all operations in detail in order to effectively control them. Changing the operations means redesigning the whole system. Contrast that with an API-based system. What matters there is only that the contract at the boundary of independent functions is respected. The independent functions implement their own controls, streamlined for consistency with interface protocols. Using interfaces, each function can contribute to a large number of very different operations at any given time.

This logic can easily be transferred to organizations. Teams organize into services around their special competence, which can then be invoked by other teams through standardized protocols that govern coordination across the whole organization. Teams can differ radically in their internal logic, tooling, and culture as long as they honor the coordination interface and the commitments made. Amazon Web Services was one of the first companies to operate and govern their operations in this way. Despite its large scale, it is still at the leading edge of innovation, and probably powering the most diverse portfolio of operations of any single company in history.

At a deeper level, the analogy of organizational interfaces is to TCP/IP, the communication standard underlying the internet since the 1980s. The internet did not require every computer to be built or operated in the same way. Different machines, operating systems, and local networks could remain radically different. What TCP/IP standardized was the interface logic of exchange: how data was addressed, transmitted, routed, and received. That made interoperability possible across highly diverse systems.

The same principle applies organizationally. Teams do not need to operate in the same way. They need to be dependable at the boundary. What this requires is surprisingly lean. A coordination architecture needs to define how teams connect, how commitments are made, how outputs are specified, how information flows, and how conflicts are handled. As long as these interface standards are respected, teams can differ internally in culture, process, and capability. Compatibility at the boundary enables freedom in the interior.

This does not eliminate the need for differentiation and integration. It changes their relationship. Integration no longer depends to the same extent on operational similarity. It can be achieved through standardized interfaces among highly differentiated units. The decades-old Lawrence and Lorsch trade-off can be laid to rest.

The Minimum Viable Governance Contract

The most important implication is that interface standards do not need to be elaborate to be effective. Over-specification is its own failure mode because it recreates the rigidity of process standardization under a different name. An interface standard needs to guarantee three non-negotiables:

• I can rely on your commitments enough to build on them.
• I can see enough of what is happening to detect problems early.
• When something breaks, there is a legitimate path to resolution.

Everything beyond these risks over-specification. The guarantees translate into a surprisingly lean governance architecture in which four practices carry most of the load.

An onboarding agreement establishes shared principles before collaboration begins. It is a clear statement of what participation requires: what is expected, what is offered, and under what conditions participants can exit. This is where the logic of the interface standard is made explicit and accepted.

Lightweight contracting specifies outputs and commitments. Traditional contracting becomes cumbersome because it tries to anticipate and specify everything in advance in response to opacity and distrust. When an interface standard exists and is respected, contracting can become more relational and adaptive. It defines what will be delivered and what happens if delivery fails, without prescribing how the work must be done.

Transparent performance figures make contributions visible without creating surveillance. Transparency supports coordination and builds trust over time without central enforcement. Teams that consistently deliver on their commitments build reputational capital. Underperformers become visible too, without anyone needing to monitor them, because they feel consequences when others are reluctant to collaborate with them.

Conflict mediation is the safety valve. When commitments fail or boundaries are disputed, there must be a trusted and accessible path to resolution that does not depend on escalation to authority. Without a clear, trusted mechanism for resolving conflicts and failed commitments, teams either escalate the issue, which reintroduces hierarchy, or avoid it, which erodes the standard over time. Mediation prevents small failures from becoming large ones.

These practices move some of the most underinvested areas of management to the center of the coordination architecture. In hierarchical organizations, they are often secondary because hierarchy carries most of the coordination load. In an interface-based coordination model, they become primary. Without them, interface-based coordination remains an idea. With them, it becomes operational.

Coordination through shared standards among autonomous contributors is not entirely new. Professional fields such as medicine, law, and academic science have long worked this way. What is new is the ability to apply this logic inside large business organizations at scale and in real time.

Technology as the Nervous System

For much of managerial history, the infrastructure needed to sustain this kind of architecture was simply unavailable. Maintaining a living, legible, and enforceable interface standard across many autonomous teams in real time required coordination overhead that consumed the gains. Today’s digital technology changes this calculus. Shared data layers, real-time performance visibility, automated workflow, and digital process management can carry much of the coordination load that previously sat on managers and meetings.

The transformation of Spanish bank BBVA illustrates how a shared digital backbone can reduce the managerial overhead of coordination by synchronizing strategies, priorities, and performance data across global operations. Critically, process users across the organization can propose improvements to the shared process architecture, and those improvements can be rapidly incorporated and made available to all operations. The interface standards themselves become more adaptive.

That is crucial. The greatest risk in any governance structure is ossification: when standards that once enabled coordination to become the bureaucracy standing in the way of flexibility. Giving participants a legitimate pathway to improve the operational system from within, “voice before exit” in Hirschman’s terms, keeps the standard connected to operational reality.

A Different Kind of Organization

What emerges is a kind of organization that is capable of complex and flexible value creation at large scale. The problem Lawrence and Lorsch identified has endured because most proposed solutions operated within the same architectural assumption: that integration requires enough operational commonality to keep differentiation under control. Standardizing coordination interfaces, but not the subjects of operation, opens a different path. It allows organizations to hold diversity and coherence together by shifting the basis of integration from shared operations to shared protocols of exchange.

What becomes possible is an organization that can scale adaptability without sacrificing coherence. It can combine deep local specialization with fluid coordination, respond to volatility without constant recentralization, and extend value creation beyond the firm into ecosystems of autonomous contributors. In that context, leadership must be reimagined. Its central task is no longer to align work through control, but to design and enable the conditions that make distributed, large-scale value creation possible.