Why are some organizations error-prone-regularly subject to interruptions and inconveniences, some of which periodically coalesce catastrophically-whereas other organizations, although similar in the products and services they generate and the process technologies they use, are reliable, adaptable, and continuously self-improving, relentlessly learning from experience to get ever better? Analyzing medical error reports and studies of high-performing, non-health care organizations reveals 2 differences. High performers know how to prevent problems from producing further consequences once they occur and how to prevent their recurrence. They do this by specifying how work is expected to proceed-who will do what for whom, with what purpose, when, where, and how-before work is actually done. Then, when anything contrary to expectations occurs, it is immediately identified as a problem. Through this approach, the effects of problems are contained, the causes are quickly investigated, process knowledge is deepened, and recurrence is prevented. In contrast, error-prone organizations tolerate ambiguity, a prevailing lack of clarity over what is supposed to happen at any given time. Problems are thus hard to identify, and, even when recognized, they are worked around. People "get the job done," but don't initiate efforts to learn from the problem or improve the process. We believe that coupling high degrees of specification with rapid responses to individual problems can improve health care. Superlative manufacturing, service, and military organizations apply this approach to myriad processes and situations, and initial health care trials of this approach have been promising. We discuss how such an approach could be initiated in health care more broadly.