We describe the integration of pattern-based reasoning learned through experience into two decision-making systems. The first is a hierarchical, multimodal game-playing program, which integrates various approaches for deciding which move to make. The second is a dynamic programming assignment system which assigns trucking resources to delivery tasks. In this case, we utilize historic patterns of activity in the network generator to limit the candidate tour generation process for the assignment algorithm. In each case, the new knowledge results in increased performance.