The key to wise decision-making in disciplines such as conservation, wildlife management, and epidemiology is the ability to predict consequences of management actions on focal systems. Predicted consequences are evaluated relative to programme objectives in order to select the favoured action. Predictions are typically based on mathematical models developed to represent hypotheses about management effects on system dynamics. For populations ranging from large mammals to plant communities to bacterial pathogens, demographic modelling is often the approach favoured for model development. State variables of such models may be population abundance, density, occupancy, or species richness, with corresponding vital rates such as rates of reproduction, survival, local extinction, and local colonisation. A key source of uncertainty that characterises such modelling efforts is the nature of relationships between management actions and vital rates. Adaptive management is a form of structured decision-making developed for decision problems that are recurrent and characterised by such structural uncertainty. One approach to incorporating this uncertainty is to base decisions on multiple models, each of which makes different predictions according to its underlying hypothesis. An information state of model weights carries information about the relative predictive abilities of the models. Monitoring of system state variables provides information about system responses, and comparison of these responses with model-based predictions provides a basis for updating the information state. Decisions emphasise the better-predicting model(s), leading to better decisions as the process proceeds. Adaptive management can thus produce optimal decisions now, while simultaneously reducing uncertainty for even better management in the future.