Teaming

A new operation made the old hierarchy dangerous

In the late 1990s, sixteen U.S. hospitals adopted a minimally invasive cardiac-surgery technology. Their cardiac departments were already excellent at conventional surgery. A familiar operation could proceed through deeply rehearsed roles: surgeon, anesthesiologist, perfusionist, and nurses often coordinated through gesture and expectation, and people within a discipline could be substituted without changing the routine.¹

The new procedure disrupted that competence. Instead of opening the chest in the conventional way, the surgical team used new equipment and an endoaortic balloon while relying on information distributed across several specialists. Actions that had once belonged mainly to the surgeon became more interdependent. A perfusionist or anesthesiologist might see a dangerous trend that the surgeon could not. Existing status relations remained, but silent execution was no longer enough.²

At the pseudonymous Mountain Medical Center, the adopting surgeon treated that rupture as a team-learning problem. He selected a small initial group for relevant skills and described trying to move from command-giver toward partner while remaining responsible for the operation. The group discussed early cases and revised its routine. Mountain ranked second on the study's implementation index. Across the high-implementation group more generally, full-team dry runs, new communication practices, and post-case reflection were common; the evidence does not establish that every high-performing site enacted every practice in an identical way.³

Chelsea Hospital began with more individual expertise. Its adopting surgeon had already performed sixty procedures elsewhere and had helped design the technology. He did not choose a special implementation team, and interviewees did not describe selection for relevant skills. Chelsea fell in the study's low-implementation group despite the surgeon's prior expertise.⁴ The contrast is the kind of episode from which Amy Edmondson builds Teaming: collective performance can depend less on assembling impressive individuals than on how people learn a new interdependence together.

The cardiac study observed implementation, not a universal team law

Edmondson, Richard Bohmer, and Gary Pisano described the sixteen hospitals in a published comparative field study. They conducted 165 interviews over five months with surgeons, anesthesiologists, nurses, perfusionists, administrators, cardiologists, and other staff.⁵ They also analyzed clinical records for 669 operations during each site's first six months with the technology. Before interpreting the interviews, they constructed a relative implementation index from procedure volume, the share of cardiac operations using the technology, and whether use was rising or falling.⁶ They compared the seven highest and seven lowest sites, leaving the two middle cases out of that classification.⁷

Across the more successful implementers, the researchers repeatedly found a sequence of enrollment, preparation, trials, and reflection. Leaders selected and motivated an initial group; teams rehearsed; people altered communication inside the operating room; and participants reviewed cases and data. The less successful sites more often staffed by availability, rehearsed partially or not at all, retained conventional communication, and rarely reflected as a whole team. The researchers developed a four-step process theory from those recurring patterns.⁸ A later analysis of the same sixteen teams reported associations between implementation, leader coaching, ease of speaking up, and boundary spanning; those relationships were observational, not estimates of causal effects.⁹

Those are observed comparative patterns. “Successful” meant that the hospital incorporated the technology into ongoing use during the study period. It did not mean superior surgery. The paper reports no difference in patient outcomes across the sites, and the index was not designed to show that adoption was clinically better than nonadoption.¹⁰ Hospitals chose to adopt before the study, leaders were not randomly assigned, and interview evidence could not eliminate unmeasured differences. The authors described the ideas as speculative and noted limits arising from the sample, retrospective interviews, imprecise measures, and the specialized setting.¹¹ The comparison generated a process theory; it did not estimate the causal effect of a four-step intervention.

The affected people also appear differently in the evidence. Clinicians described status, voice, preparation, and frustration; patients supplied the stakes and clinical records but were not among the interview groups the authors list.⁵ A team can become excellent at implementing a technology without the study settling who should receive it, how consent should work, or whether its benefits exceed its risks.

From learning a routine to forming collaboration in motion

Published in 2012, Teaming joins this and other research with cases from several kinds of organizations. Its central move is grammatical. A “team” sounds like a bounded, stable unit; teaming names the activity of coordinating and learning while participants are already acting, often without stable team structures.¹²

The cardiac case gives the generalization both support and friction. Mountain did not begin with a completely fluid coalition. Early membership stability helped people build a new routine, and the researchers explicitly warned that rotating newcomers too quickly could interrupt learning. Yet keeping a special team closed for too long could create resentment and prevent the practice from spreading. The coalition therefore had to change over time: first concentrated enough to learn, then permeable enough to diffuse what it had learned. Teaming does not make stability obsolete; it asks leaders to match boundaries and membership to the stage and uncertainty of the work.¹³

Psychological safety is one condition in that process. People need to ask for help, admit uncertainty, and interrupt when their information matters. At Mountain, that meant changing the real consequence of a lower-status clinician speaking to a surgeon, not merely announcing that candor was welcome; the later analysis found ease of speaking up associated with implementation but did not establish causation.⁹ This connects the book to Organizational Learning II, where defensive routines protect competent appearances, and to The Fearless Organization, which develops the voice construct for a wider audience.

Boundary spanning and experimentation carry the argument beyond voice. Groups need representations and language through which specialized knowledge can meet without pretending that every participant knows the same thing. They also need short cycles in which action produces interpretable feedback. The book treats framing, safety, learning behavior, and boundary crossing as mutually reinforcing practices.¹⁴ The hospital studies measured narrower variables and processes, not a single “teaming” treatment. A 2022 review applied the concept to coordination in fertility clinics while noting that teaming had a limited foundation in that clinical setting—evidence of later uptake and of a remaining validation gap, not proof of the whole framework.¹⁵

The Chile rescue became a retrospective illustration

The 2010 rescue of thirty-three miners in Chile already appeared in Edmondson's 2012 account of teaming. Her later work with Jean-François Harvey recast it as “extreme teaming.” In both uses, the case is a retrospective illustration, not a controlled test that the concept caused the rescue.¹⁶

In Edmondson and Harvey's telling, André Sougarret organized parallel efforts; failed drilling attempts became feedback; and Felipe Matthews took on a measurement role after testing equipment at the site.¹⁶ NASA's participant record supports a narrower part of the account: its advisers helped develop capsule requirements and worked with Chilean engineers and the Chilean Navy, while emphasizing that Chileans designed and built the capsule.¹⁷ All thirty-three miners emerged after sixty-nine days.¹⁸ The episode depicts a changing coalition learning through bounded trials, but the sources do not isolate “extreme teaming” as the cause of success.

The rescue narrative should not erase the conditions that placed the workers underground. A Chilean Chamber of Deputies investigative commission assigned responsibility to the mine owners and the national mining-safety regulator, including failures of safety compliance and oversight around reopening the mine.¹⁹

That limit poses questions the framework does not resolve. Who performs the translation, relationship, and context-switching labor of fluid collaboration? Do they have protection, authority, time, and recognition? When experiments affect patients, workers, users, or communities, who sets the boundaries and represents those bearing the risk? Teaming is most useful when its practices widen who can shape the work and what the organization can learn; the concept should not turn flexibility itself into evidence of benefit.

Structured reading paths and evidence limits

The Amy C. Edmondson path identifies the author and locates the work within her research program. Organizational Learning II supplies an important conceptual predecessor for its treatment of defensive routines, while The Fearless Organization develops Edmondson's later public synthesis of psychological safety. Team of Teams is an editorial comparison between two accounts of coordination under uncertainty, not evidence that either work supplied the other's method or findings.

The coordination, communication, and common understanding path connects the book's boundary-spanning practices to a broader coordination problem. The culture, informal organization, trust, and voice path highlights the status relations and informal consequences that determine whether speaking up is actually safe. Organizational intelligence gathers those learning and coordination mechanisms into a wider organizational capability; benefit for all life supplies the moral boundary for asking whether adaptation benefits the patients, workers, communities, and living systems exposed to it.

No structured impacts, typed relations, or formal reading dependencies are asserted for this work. The related paths above are editorial navigation; they do not by themselves establish empirical support, intellectual influence, or a shared method. The evidence package combines a comparative hospital field study, a later analysis of the same teams, Edmondson's primary synthesis, bounded clinical uptake, retrospective author interpretation of the Chile rescue, participant and agency records, and an official legislative investigation. It does not include representative testimony from patients, miners or their families, lower-status clinicians, contingent workers, or affected communities, nor does it measure ecological outcomes.²⁰