To enable adaptive training for Soldiers, Sailors, and Airmen in simulations and immersive environments, adaptive training systems must become affordable. Currently, they require handcrafting by experts with specialized knowledge of Artificial Intelligence working closely with domain experts. We focus on three complementary tasks that address affordable adaptive tutoring: (1) identifying performance weaknesses on which trainees must improve; (2) making better tutoring decisions; and (3) integrating adaptive instruction into a common architecture enabling capabilities developed once to be applied elsewhere. We report experimental results from studies investigating these tasks. We have applied expert policy capture to identify student performance weaknesses. Experts (1) reviewed many students’ performance in a simulation, (2) critiqued that performance, and (3) rated performance for overall quality. Analysts constructed scoring rules that mimicked experts’ judgments. For each identified area of student weakness, an instructional intervention was constructed. To determine which instructional interventions students should interact with, we are using a modular reinforcement learning framework that uses the history of many learners to determine which instructional approaches work best for which types of learners and contexts. This approach decomposes adaptive instruction in terms of adaptable event sequences, which are each separately modeled as a Markov decision process (MDP). The instructional policies for the MDPs are induced using off-line reinforcement learning techniques with a corpus of prior learner interaction data. The instruction is presented within a unifying framework for adaptive instruction called the Generalized Intelligent Framework for Tutoring (GIFT). GIFT enables instructional capabilities that are developed within GIFT to be applied in the future to other instructional systems and training environments. GIFT authoring tools support domain independent principles of instruction, as well as domain specific instruction.