The capability to provide on-demand service access for SMEs can further reduce costs and allow companies to concentrate investments on their core businesses which in turn facilitates the overall competitive advantage. In the FP7 European Project COIN, a platform for supporting SMEs is developed which combines the flexibility of an execution environment for Semantic Web services with agent-based service compositions. This paper presents an approach to integrate an implementation of a Semantic Web service platform with an agent platform. One major benefit of the combination is that through this the potential of agent systems, most notably flexibility, can be tapped. An obstacle is the integration of dynamically discovered services which often cannot be used by agents because of interoperability problems (e.g. data heterogeneity). By delegating this task to a Semantically Enabled Service Oriented Architecture (SESA), the agent platform can concentrate on coordination tasks. Coordination tasks include executing predefined compositions of services but also automatic compositions of semantic services by applying AI planning techniques, in particular by transforming service descriptions into the Planning Domain Definition Language (PDDL). The integration approach is motivated and described in the context of a high-level scenario coming from the area of enterprise interoperability.