Most PM systems are intended to serve in a hybrid machine {human environment [1]. Thus, variability through customization and evolution is crucial. As for interactive and knowledge-based systems, many PM systems have utilized interpretation and re ection, often combined with object-orientation. However, there is meagre methodological support for controlling variability eg of a Process Schema and its instances. The paper will rst de ne the term process model. Then follows a list of demands and existing technologies for process model variability. We also characterize some existing PM systems wrt. to such variability. Thereafter follows a short presentation of the EPOS support for PM variability, and some indications of further work.

The EPOS approach to PM variability consists of: 1) A common, re exive, object-oriented datamodel, available through the SPELL process specication language and o ering types to describe the software process world; 2) an AI Planner to incrementally (re) generate task networks; 3) an underlying EPOSDB versioned database that stores both the Process Schema and its model instances.