The high cost of live training has always been a major challenge for the military. This challenge will only grow as current fiscal uncertainty leads to declining training budgets. Constructive simulations, such as One Semi-Automated Forces (OneSAF), have shown to partially reduce some costs associated with warfighter training. However, further cost reductions in simulation are always sought to ensure that simulation remains an attractive training option for the Commander. The Army Research Laboratory-Human Research and Engineering Directorate, Simulation and Training Technology Center focused on an effort to lessen costs by creating an automatic 'driver' for OneSAF with less need for human intervention. For this initial effort, we describe how the Linguistic Geometry Real-Time Adversarial Intelligence and Decision-making (LG-RAID) lightweight simulation generated and sent to OneSAF tactically valid cooperative entity behaviors for an entire company-size-force of friendly and enemy combatants. We discuss how this was accomplished, for both scenario creation as well as scenario execution. For this initial paper, our results primarily focused on scenario creation, with follow-on studies concentrating on scenario execution. This paper describes key principles developed behind the 'driver' and offers potential areas for future research based upon our lessons learned in this study.