Nowadays, the use of Virtual Reality for training and educating people has become widespread. Using simulators for training has many possible advantages such as increased safety, reduced costs and the fact that a conditioned training environment can be used. Currently, most researchers in the field are focusing on improving simulator realism and corresponding hardware requirements. However, not much attention has been paid yet to the automation of the training process. The current generation of training simulators is not sophisticated enough to fully replace a human instructor. It is hypothesized that a computer is better able to train a human being than a human instructor would, as a computer can objectively measure performance and automatically keep track of the student's skills on multiple tasks. Currently Delft University of Technology and Green Dino are developing a so-called Virtual Assistant (VA). The VA analyses human behavior and its operating performance (1. analysis), adapts to the student (2. adaptation), and efficiently communicates with the student (3. communication). Analyzing human behavior deals with measuring and judging performance from event-related data and continuous signals. Generic methods are being developed to quantify human performance from the complexity and frequency of movements. Adaptation deals with creating a student profile according to the student's short-term and long-term goals, needs and capabilities, such that optimal training is supplied. Communication deals with improving efficiency of communication by possibly using other modalities such as visual and haptic displays, rather than speech alone as used by human instructors. In the future, the VA should be able to fully replace a human instructor, and to supply more efficient and more effective training.