Accepted practice absolves building energy modelers of responsibility for capturing many of the effects of occupant behavior by assuming fixed comfort targets and ignoring “unregulated” loads. This paper asks what we can learn by incorporating more detailed information about occupant behavior into models. It compares results of three approaches: conventional practice, an augmentation incorporating detailed occupancy patterns, and an augmentation incorporating detailed behavioral responses of occupants to evolving comfort conditions. We apply these models to a highly-instrumented commercial building in Philadelphia, PA, USA, using EnergyPlus and extensions based in Markov chain modeling and agent-based modeling. We share preliminary findings only because the project schedule was disrupted. Key preliminary findings are that (1) better occupancy data greatly improves energy model accuracy, (2) standard assumptions about occupant schedules are often wrong so that a more sophisticated representation is warranted, (3) better data about occupants’ adaptive responses only marginally improves energy model accuracy, (4) yet such data are quite valuable for predicting occupant satisfaction, and (5) incorporating occupancy data EnergyPlus needs additional hooks for incorporating occupant behavior.