We believe that computer animation in the form of narrated animated simulations can provide an engaging, effective and flexible medium for instructing agents in the performance of tasks. However, we argue that the only way to achieve the kind of flexibility needed to instruct agents of varying capabilities to perform tasks with varying demands in work places of varying layout is to drive both animation and narration from a common representation that embodies the same conceptualization of tasks and actions as Natural Language itself. To this end, we are exploring the use of Natural Language instructions to drive animated simulations. In this paper, we discuss the relationship between instructions and behavior that underlie our work and the overall structure of our system. We then describe in some what more detail three aspects of the system - the representation used by the Simulator, the operation of the Simulator and the Motion Generators used in the system.