During scenario-based training, the scenario is dynamically adapted in real time to control the storyline and increase its effectiveness. A team of experienced staff members is required to manage and perform the adaptations. They manipulate the storyline and the level of support during their role-play of important characters in the scenario. The costs of training could be reduced if the adaptation is automated by using intelligent agent technology to control the characters within a virtual training environment (a serious game). However, such a system also needs a didactical component to monitor the trainee and determine necessary adaptations to the scenario. This paper investigates the automation of didactical knowledge and the corresponding dynamic adaptation of the scenario. A so-called director decides upon and distributes the necessary changes in real-time to the characters. First, the nature and goals of the adaptations are analyzed. Subsequently, the paper introduces a conducted study into the applicability of directable scenarios. Thereafter, an experiment is introduced that investigates the effects of directorial interventions upon the instructive quality of the scenario. Qualitative results indicated that trainees experienced scenario-based training to be instructive and motivating. Moreover, quantitative results showed that instructors rated directed scenarios as significantly better attuned to the trainee’s needs compared to non-directed scenarios. Our future research will focus at the design of an architecture for automatically directed scenario-based training. © 2013 Elsevier Ltd. All rights reserved.