Using AI and semantic web technologies to attack process complexity in open systems

Recently many vendors and groups have advocated using BPEL and WS-BPEL as a workflow language to encapsulate business logic. While encapsulating workflow and process logic in one place is a sensible architectural decision, the implementation of complex workflows suffers from the same problems that made managing and maintaining hierarchical procedural programs difficult. BPEL lacks constructs for logical modularity such as the requirements construct from the STL [STL 2003, Introduction to the STL. Available from: .] or the ability to adapt constructs like pure abstract classes for the same purpose. We describe a system that uses semantic web and agent concepts to implement an abstraction layer for BPEL based on the notion of Goals and service typing. AI planning was used to enable process engineers to create and validate systems that used services and goals as first class concepts and compiled processes at run time for execution.