Misspecification of the nature of organizations may be a major reason for difficulty in achieving performance improvement. Organizations are often viewed as machine-like, but complexity science suggests that organizations should be viewed as complex adaptive systems. I identify the characteristics of complex adaptive systems and give examples of management errors that may be made when these characteristics are ignored. Command, control and planning are presented as managerial tasks that come to the fore when a machine view of organizations dominates thinking. When we treat organizations as complex adaptive systems the focus of managerial activity changes, and sensemaking, learning and improvisation become appropriate strategies for performance improvement. Each of these is defined and described. A modest research agenda is presented.