There are many challenging problems facing the system designer in the application of the automated speech technologies to simulation. We have attempted to show that effective strategies can be designed to employ these technologies in a training environment. These strategies include techniques for augmenting shortcomings both in the state of the art and in user performance to produce a system which can respond intelligently, be acceptable to the user, and can provide excellent training. These concepts will soon be ready for testing, and we look forward to a successful field evaluation.