Designing the adequate scope and granularity of services is critical for their effective reuse. Patterns at business process level are abstractions of common and reusable designs to operate businesses. Business Process (BP) patterns can capture expert process design knowledge and greatly benefit the design of new enterprise services by guiding the definition of their scope and granularity. Identifying pattern instances in real and large documented business processes is a challenging task, requiring the analysis of the structure, semantics and behaviour associated to process descriptions. In this paper we present a solution to identify BP patterns based on a graph matching mechanism. Structural and semantics aspects, including natural language processing, are addressed. The approach moves one step further to increase automation during the design of processcentric enterprise services. We demonstrate the approach, discuss its limitations, novelty and practical benefits by using a case study based on the National Revenue Agency case at SOPOSE08