While ensuring safety; health; and comfort are indispensable, today's challenges in building control frequently hinge around systems integration and reducing, predicting, and flattening site energy consumption. To achieve these goals, it is likely necessary that we utilize new perspectives, approaches, and techniques made possible by the rapidly growing digital world. This paper highlights a process that incorporates a handful of these tools, providing basic background to their implementation. Within the multiagent systems (MAS) paradigm, the structure incorporates probabilistic graphical modeling techniques and conditional satisficing games in a way that is intended to facilitate more autonomous control capabilities. The applied theory will be used to give individual equipment controllers a framework to explicitly consider uncertainty in control actions, to simultaneously seek high performance individually and as a system of equipment, and permit autonomous adaptability. The structure may eventually allow for the improvement of ad hoc, heuristic strategies with statistical machine learning techniques, all while building a system that fundamentally incorporates a system-wide perspective of operation.