A key component of realistic and effective training in simulation is the behavior of semi-automated forces (SAFs). SAFs provide opponents, friendly forces, and other dynamic entities within the simulation. In most cases today, SAFs are designed and implemented to be tactically realistic; that is, they take actions that carry out good tactical decisions. As a result, SAFs are typically evaluated in terms of the realism or “fidelity� of their actions to the tactical situation and not with regard to training effectiveness. We contend SAF tactical realism is a necessary but incomplete requirement for cost-effective and training-effective deployment of SAFs for simulation-based training. SAF behavior should also be modulated by scenario/exercise goals and also by the learning needs of individual trainees. In practice, these additional requirements tend to surface during delivery of training, requiring human instructor/operator teams to intervene. Interventions both increase the cost of simulation-based training and potentially lower the aggregate effectiveness of that training: delivering an appropriate experience at an apt time to the trainee is contingent on the attention and action of the instructional team. Further, as SAFs are increasingly used in mixed live-virtual-constructive training situations, SAFs that consider only tactical decisions will further limit scalability and increase the operational cost of LVC training. In response, we suggest that imbuing the training system with the capability to understand and support scenario goals and individual training needs can make SAFs more practical for everyday training. We present examples of adaptation and variation that may be important for training but that are not typically embedded in a tactical SAF. We discuss the implications of these missed requirements and outline suggestions for incorporating interpretations of learning context in future simulation systems based on experience researching and developing such a capability. Finally, we outline methods for verifying and validating SAFs designed to meet these additional requirements.