The volumes of data in information systems are growing drastically. The systems become increasingly complex in trying to handle heterogeneity of ubiquitous components, standards, data formats, etc. According to the vision of Autonomic Computing, the complexity can be handled by introducing self-manageable components able to "run themselves." Agent Technology fits this vision, whereas interoperability among autonomic components can be tackled by Semantic Technologies. The problem of efficient heterogeneous data sharing, exchange and reuse within such systems plays a key role. We present an approach of constructing semantic capabilities (self-descriptive functional components) for software agents and a mechanism for distributed data management that applies these capabilities to build various industrial business intelligence systems.