Using Large Language Models in Business Processes

Large language models, such as ChatGPT, provide ample opportunities for organizational work. These models are capable of collecting, integrating, and generating information with no or little human supervision [1]. Despite their wide and rapid uptake, we lack systematic knowledge about how large language models can be used in business processes. Our tutorial sheds light on the organizational, managerial and design-related implications of using large language models in business processes. We present a theoretical framework that integrates and synthesizes research from relevant streams, including task complexity [2], task automation [5], and human-AI delegation [1]. We specify potential opportunities and threats in relation to various forms of tasks, such as decision tasks and judgment tasks. Along these lines, we also explore how the use of large language models may affect the overall outcome of a process, for example, by providing new value propositions. We use, reflect, and discuss the implications of our framework based on real-world examples. Our conceptual framework is relevant to guide future research [e.g. 1] but also inform managerial decisions in organizations [e.g. 3].