Scholarlink

The compliance in business processes is vital for maintaining smooth enterprise operations. Traditional compliance risk identification solutions heavily rely on manual inspections and expert analysis, which is time-consuming, labor-intensive, and inflexible when adapting to specification modifications. Inspired by achievements of Large Language Models (LLMs), we are exploring a novel approach to enhancing compliance risk management. In this paper, we propose an automated framework for risk identification in business processes using LLMs, named AFPCRL. AFPCRL first utilizes prompts to guide LLMs in interpreting and structuring lengthy process management specifications into coherent process triple sequences. Besides, AFPCRL introduces a novel compliance assessment mechanism designed to detect discrepancies between generated standard processes and actual business processes, providing an effective strategy for identifying compliance risk points. Experiments conducted on real business scenarios demonstrate that AFPCRL significantly outperforms existing baselines, underscoring the efficacy and potential of AFPCRL in expending LLMs for business process risk identification.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Leveraging LLMs for Automated Compliance Risk Identification in Business Processes

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Yue Wang,
  • Ningyuan Yi,
  • Wenjing Chang,
  • Jianjun Yu

摘要

The compliance in business processes is vital for maintaining smooth enterprise operations. Traditional compliance risk identification solutions heavily rely on manual inspections and expert analysis, which is time-consuming, labor-intensive, and inflexible when adapting to specification modifications. Inspired by achievements of Large Language Models (LLMs), we are exploring a novel approach to enhancing compliance risk management. In this paper, we propose an automated framework for risk identification in business processes using LLMs, named AFPCRL. AFPCRL first utilizes prompts to guide LLMs in interpreting and structuring lengthy process management specifications into coherent process triple sequences. Besides, AFPCRL introduces a novel compliance assessment mechanism designed to detect discrepancies between generated standard processes and actual business processes, providing an effective strategy for identifying compliance risk points. Experiments conducted on real business scenarios demonstrate that AFPCRL significantly outperforms existing baselines, underscoring the efficacy and potential of AFPCRL in expending LLMs for business process risk identification.