Today's dynamic and uncertain business environment requires quick reaction to change and frequent deviations from plans, making business agility indispensable. Therefore process-based systems must be able to flexibly adapt to change and provide learning capabilities. This paper proposes an approach to achieve agility in workflow management systems based on the integration of workflow management and conversational case-based reasoning. Keeping semantical knowledge about ad-hoc deviations in cases allows starting with a lightweight initial workflow model and fosters learning from living processes to continuously improve workflow execution.