LLMs and Process Mining: Challenges in RPA

Process mining is often used to identify opportunities for process automation leading to improved efficiency and cost savings. Robotic process automation (RPA) is a fast-growing area that provides tremendous productivity growth to a growing number of companies across many industries. RPA tools allow users to record their work and then propose areas for automation, and produce scripts to automate work. Recording how a process is conducted, coupled with process mining techniques, offers the most detailed view of what process is followed, how well it is followed, and whether there are areas for automation or improvement in process or policies. However, the main challenge to deriving these insights is the need for grouping fine-grained recorded tasks into events, giving them names, and proposing how to automate those high-level tasks. In this paper, we propose a framework for using large language models (LLMs) to assist users in these steps, by leveraging their natural language understanding and generation capabilities. We first address the problem of event log generation, which is an input of automation techniques, by using LLMs to group and label tasks based on their semantic similarity and context. We then tackle the problem of connector recommendation by using LLMs to recommend best plugins to automate tasks. We evaluate our approach on a real publicly available dataset, and show that it can improve the quality and efficiency of event log generation and connector recommendation, compared to the baseline methods.