The increasingly daunting problem of antimicrobial resistance has led to an intense focus on optimization of antibiotic therapy, with simultaneous goals of improving patient outcomes and minimizing the contribution of that therapy to making the available antibiotics obsolete. Although even appropriate antibiotic therapy drives resistance, inappropriate therapy may also have adverse effects on the individual patient, as well as on the bacterial ecology. Recent research has validated the benefit of intelligent utilization of both microbiological data and clinical assessment in the empirical selection of initial broad-spectrum therapy and in further guidance of therapeutic decisions throughout the course of illness by use of a systems approach. Thus, the optimal approach to the critically ill patient with infection involves the initiation of aggressive broad-spectrum empirical therapy followed by timely responses to microbiological and clinical results as they become available. An appropriate response to this information often involves de-escalation of therapy or even its discontinuation.