The essence of Case-Based Reasoning (CBR) as a problem solving paradigm is that solutions are generated by adapting the solutions of similar problems rather than solving the problem from first principles. In this paper we present a categorisation of problem solving tasks, arranged according to complexity. In addition we categorise CBR systems according to the complexity of the adaptation process involved. We describe three CBR systems; a system for property valuation, a system for software design and a system for modelling in engineering analysis. We discuss the manner in which the advantage of a CBR solution to these problems shifts as the task becomes more complex and the complexity of the adaptation process changes.