In 2002 and 2003, five government, scholastic, and industry research groups worked together to attempt the AAAI Robot Challenge. In short, autonomous robots are to attempt the roles of graduate students at the American Association of Artificial Intelligence (AAAI) conference. Two of the main thrusts of the Robot Challenge are for robots to demonstrate autonomy and to interact with people in natural and dynamic environments. This paper focuses on the task of the robot finding its way from the front door of the conference center to the registration desk by asking for directions. Asking for and following route directions in a partially known environment provides a rich testbed for interacting with an autonomous system. This paper describes the task and the environment, details the adjustable autonomy architecture implemented for the Robot Challenge, and compares it against other approaches. It also presents the problems encountered using this architecture, and suggests further refinements.