Middleware support for business process management (BPM) has met some of the challenges with respect to encoding, performance and maintenance of workflows. A remaining challenge is complexity: business processes are becoming widely distributed, interoperating across a range of inter- and intra-organizational behaviours, vocabularies and semantics. It is important that this semantic complexity is checked and analyzed for optimality and trustworthiness prior to deployment. Petri nets are a??formal method that successfully provides behavioural analysis. A shortcoming of Petri nets is that the data exchanged between business activities abstract too far away from the importance of data in actual business processes. This paper addresses this abstraction gap via additional semantic enrichment, through a two stage, model-driven approach.