Accelerating Model-Based Systems Engineering by Harnessing Generative AI

With the rise of artificial intelligence (AI) tools to support the work of numerous disciplines, we describe a preliminary investigation into the benefits and drawbacks of large language model (LLM) use as part of a traditional systems engineering and design workflow. To explore this, we tasked a group of systems engineers to each create a list of requirements and use case diagram to satisfy a systems of systems user scenario presented in a proposal document. Participants created models of a healthcare setting in which clinicians resolved discrepancies with patient care by consulting additional sources of record, demonstrating the importance of integrating new systems within the larger healthcare system of systems. The first group were provided open access to an LLM, the second group were provided draft materials generated by an LLM, and the third followed their normal workflow. A subject matter expert (SME) evaluator then scored each model according to its completeness, consistency, correctness, simplicity, and traceability. Through this, we show that although LLMs are not a replacement for a trained systems engineer, they can contribute in two primary ways to the modeling process: first, they can generate a significant portion of the information necessary to create a minimum viable product (MVP) model within a fraction of the time, offering a promising way to accelerate the overall model development process. Second, they can answer detailed, domain-specific questions and reduce the time spent on external research.