The primary goal of our research is to build an intelligent tutoring system for red blood cell antibody identification. In this paper, we describe the basis for a tutoring system--an expert system called RedSoar. RedSoar is built from two task-specific architectures that were designed for building flexible systems (i.e. systems that can use a variety of problem-solving strategies, and to which knowledge can easily-be added). RedSoar solves the antibody identification task correctly 81% of the time and new knowledge can be added in a straightforward manner. The system is capable of exhibiting human-like behavior which we believe is a necessary condition for building a successful tutoring system.