Automatic preferences can influence a decision maker’s choice before any relevant or meaningful information is available. The authors account for this element of human cognition in a computational model of problem solving that involves active trial and error and show that automatic biases are not just a beneficial or detrimental property: they are a tool that, if properly managed over time, can give rise to superior performance. In particular, automatic preferences are beneficial early on and detrimental at later stages. What is more, additional value can be generated by a timely rebiasing, i.e. a calculated reversal of the initial automatic preference. Remarkably, rebiasing can dominate not only debiasing (i.e., eliminating the bias) but also continuously unbiased decision making. This research contributes to the debate on the adaptiveness of automatic and intuitive biases, which has centered primarily on one-shot controlled laboratory experiments, by simulating outcomes across extended time spans. The authors also illustrate the value of the novel intervention of adopting the opposite automatic preference - something organizations can readily achieve by changing key decision makers - as opposed to attempting to correct for or simply accepting the ubiquity of such biases.