Our research is exploring the impact of the role(s) that people will play in large-scale multi-agent systems. We must develop strategies for coordinating not only the problem solving behavior of autonomous agent communities, but also their information sharing and mixed-initiative human interaction behavior. In this paper, we describe our experience combining two interactive agent systems: TRIPS, The Rochester Interactive Planning System [3, 4], and CAMPS-MP, an interactive airlift scheduling tool developed for the Air Force [2]. This revealed requirements for effective multi-agent mixed-initiative interactions, including the role of explanation and the need for contextual information sharing among the agents.