LLM-based RPA System for Optimizing Workflows in Financial Risk Management

his paper aims to leverage the advancements in General Computer Control (GCC) to improve the efficiency and effectiveness of risk management operations in financial institutions. Specifically, we introduce an LLM-based Robotic Process Automation (RPA) framework designed to enhance front-line employee work, adapt to the specific needs of financial institutions, and automate tasks requiring minimal cognitive effort. To demonstrate the effectiveness of our proposed framework, stress testing, a common task for risk management department, is used as a case study. The results show that the RPA system can improve efficiency, reduce costs, and minimize errors, all without significantly altering the existing workflow. Moreover, to address customer information security and prompts copyright protection issues, a storage method that separates the server from the client is used. Finally, empirical evidence implies that even models with weaker capabilities can achieve the desired work objectives when guided by detailed prompts.