Among the many anticipated roles for robots in future is that of being a human teammate. Aside from all the technological hurdles that have to be overcome on the hardware and control sides to make robots fit for work with humans, the added complication here is that humans have many conscious and subconscious expectations of their teammates – indeed, teaming is mostly a cognitive rather than physical coordination activity. This focus on cognitive coordination, however, introduces new challenges for the robotics community that require fundamental changes to the traditional view of autonomous agents. In this paper, we provide an analysis of the differences between traditional autonomous robots and robots that team with humans, identifying the necessary teaming capabilities that are largely missing from current robotic systems. We then focus on the important challenges that are unique and of particular importance to human-robot teaming, especially from the point of view of the deliberative process of the autonomous agent, and sketch potential ways to address them.