Almost 20 years ago, the vision of using enterprise models not only for analysis and design phases, but also throughout the entire life cycle of corresponding enterprise software, emerged at the chair. A concretisation of this vision was described in 2008 by Ulrich Frank and Stefan Strecker in a research report (download).

The label ‘self-referential enterprise system’ was intended to express that enterprise software is integrated with a corresponding enterprise model and can therefore refer not only to its conceptual basis, but also to the context in which it is used. Such an architecture would offer users the opportunity to navigate an enterprise model during runtime and, if necessary, make changes there that affect the enterprise software. However, the analysis of possible implementation approaches showed that a close integration of enterprise software and enterprise model was not possible, which was mainly due to fundamental limitations of the programming languages available at the time. With the development of the multi-level language architecture based on XModeler, which began in 2010, it has now been possible for several years to realise a convincing implementation of self-referential enterprise software. The multi-level language architecture allows the joint representation of models and programmes. It is therefore no longer necessary to synchronise two separate representations. Language and meta-language levels are also integrated into the architecture. Authorised users can therefore not only adapt models (and, hence, code) during runtime, but also the corresponding (modelling and programming) languages if required. A current description of the architecture and the associated promises can be found in the paper ‘Multi-Level Language Architectures as a Foundation for Advanced Enterprise Systems’ (download).