Enterprise performance is, now more than ever, one of the key points for reaching the market success. In order to increase it, economics paradigms focus on how to better manage knowledge acquiring, sharing and update. Knowledge management can be approached with the possibility offered by the sustainability goals trying to optimize different enterprise strategic domains. The modern architecture of information systems (ISs) is based on distributed networks with a grand challenge of representing and sharing knowledge managed by ISs and consequently, to remove semantics interoperability barriers. First, this paper analyses interoperability issues between Cooperative Enterprise Information Systems (CEIS). Based on this analysis, the authors propose a conceptualisation approach for semantics discovery and management in Enterprise Information Systems models, based on applying fact-oriented transformation rules. The input of the transformation process is a conceptualised UML class model, reverse-engineered from an implemented model, and transformed into a Fact-Oriented Model (FOM), which makes explicit the finest-grained semantics. Semantics aggregates are then computed for structuring the whole semantics embedded in enterprise applications. They define independent set of concepts with their own minimal mandatory semantics. Finally a case study is proposed to validate the practicability of our approach in a real scaled scenario involving an Enterprise Resource Planning (ERP) and a Manufacturing Execution System (MES).