This paper presents a pragmatic knowledge-based approach to evolving and reengineering legacy business software systems. We describe a CASE tool for assisting legacy system evolution by automating the tedious and knowledge-intensive conversion process and show how we developed and maintained the tool and how it is used with payoff. This work demonstrates that timely technology upgrades of legacy systems are only possible with carefully engineered knowledge-based tools. Viewing software as an artifact with structure, behavior, and function, we can represent most program conversion knowledge explicitly for function-preserving transformation in an automated conversion system. The payoff for using the knowledge-based approach to software evolution is not only in terms of valuable resources saved, but also in terms of improved quality of the upgraded software systems.