Kurt Sandkuhl, Balbir Barn, Souvik Barat
Abstract Large language models (LLMs) are considered by many researchers to be a promising technology for automating routine tasks in conceptual modeling, such as in enterprise modeling. Enterprise Modelling (EM) is a significant method-based endeavour that is often seen as a prerequisite to broader organisational programmes of work, such as digital transformation. Examples of tasks in EM that could be supported by LLMs include generating models from natural language descriptions provided by domain experts and transforming models into text to support non-modeling experts in applying models. For such tasks, it would be beneficial to ensure consistent terminology use across enterprise models, LLM prompts, and natural language descriptions. The research question addressed in this paper is: In the context of enterprise modelling, what is required to conceptually integrate prompt and enterprise modelling meta-models? From this exploration, the main contributions are (1) a novel and first such example of a conceptual meta-model for prompt engineering that integrates the concepts of the modelling domain under consideration with concepts from the modelling language applied, and the input and output of prompts, and (2) a demonstration of the applicability of the meta-model for different EM languages. The application of this meta-model, created using design science principles, is examined through an evaluation strategy comprising three evaluation episodes: internal to development team, external in the validation context and external in the application context. The proposed meta-model is potentially significant as it paves a route for the design for prompt modelling tools in an enterprise modelling context.