Increasingly business are looking to take advantage of data driven techniques and technologies e.g. analytics, machine learning, deep learning, AI. The biggest blocker to the successful application of these data driven techniques is problems with the data itself. Data is most often created in isolated silos across the organisation. Many different design patterns for data integration have been developed over recent years to try to overcome the challenges this slioed data presents. Though the approaches vary, the thing they have most in common is a high rate of failure. Every business has a story of ERP, MDM and data-warehouse projects that have failed to deliver. What these failed approaches have most in common is the requirement for the data to conform to a single model. In this talk we will demonstrate that by adopting a multi-model approach with semantics at the core the chances of succeeding can be greatly improved.