This contribution discusses the well-known paradox of today's approaches to standardizing of semantic models in automation engineering and proposes an evolutionary concept to resolve it. In this context, an overview of past and current standardization activities is presented which points out the existing barriers. Instead of aiming for a completely standardized neutral data model, the proposed approach actively supports data exchange with mixed neutral and proprietary data models. Hence, it utilizes the existing heterogeneity and the maturity of proprietary data models. In this context, the authors present a concept of maturity levels, which form milestones towards a stepwise evolution of a semantic standardization. Finally, this paper describes how this approach is technically implemented in AutomationML.