Improving the efficiency of training complex cognitive tasks necessitates controlling cognitive load during learning. Adaptive training manages the cognitive load by adjusting the training schedule and difficulty to match immediate performance. This experiment is innovative in both the complexity of the experimental task and the adaptation scheme. The complex cognitive planning task required participants to choose and route unmanned vehicles to reconnoiter geographic areas. Each scenario had multiple suitable solutions and there were many irrelevant, relevant, and key factors that could be considered with regards to the vehicles, terrain, and mission. Adaptation was based on both performance and decision process (i.e., number of factors considered). We compared two types of adaptation to a typical graduated-difficulty condition and a constant control. When all training conditions were combined, there was a significant correlation between key factors and performance. The more key factors considered, the better the performance. Adapting by decision process (as opposed to adapting by performance alone) had a marginal effect on performance. For those that started with low difficulty, those that adapted by decision process made better plans than those that adapted by performance alone. This study provides evidence on the usefulness of adapting by decision process and warrants further research on the topic.