Over the past year, Large Language Models (LLMs) have demonstrated their value across diverse industrial applications, from supporting software development to automating customer interactions and enhancing process automation. This paper explores the potential of LLMs in constructing Enterprise Digital Twins (EDTs), an emerging decision-making tool for various business sectors. EDTs facilitate “in silico” business experimentation, enabling evidence-based decision-making; however, their construction demands deep domain expertise across multiple stakeholders and seamless coordination between domain experts and expert modelers, making the process highly challenging. This study investigates the role of LLMs as domain experts and modelers to reduce dependencies on specialized expertise and their coordination. Our approach integrates meta-modeling and Model-Driven Engineering (MDE) techniques, leveraging LLMs to alleviate the cognitive burden on domain experts while employing a systematic, metamodel-guided method for creating purposive digital twins. The efficacy of this approach is illustrated using a real-life EDT use case.