Enterprise modelling is one of the popular means for capturing organizational knowledge in representations that are both human-readable (that is, diagrammatic) and machine-readable (that is, sufficiently formal and granular). However, the discipline faces certain challenges pertaining to the complexity of the socio-technical system to be modelled. This often requires a separation of concerns, to allow separate modelling of various enterprise facets – work processes, organizational structure, resource descriptions. This separation must be compensated by consistency management approaches, as diagrams of different types become inter-dependent views, enabled by different viewpoints (modelling language fragments with different scopes). The paper introduces novel means for managing the consistency of knowledge captured in multi-view enterprise models. The proposal is based on semantic graphs derived from diagrammatic models, and queries acting upon these. To achieve the goal, the Linked Data paradigm is repurposed and its technological enablers are aligned to the underlying graph nature of enterprise models. Several use cases will be discussed: (a) view transformations through graph rewriting; (b) view synchronization through reasoning; (c) passive view consistency checks. Exemplary cases are extracted from two research projects where the proposal has been successfully applied, therefore background on the projects will be provided to facilitate understanding.