From Data to Decisions: Enhancing Loan Applications with Process Mining Techniques

Abstract

This research investigates the loan application process using Process Mining techniques combined with Generative AI to identify inefficiencies and propose improvements. Data from the BPI Challenge 2017 was analyzed using Process Mining tools, particularly the Fuzzy Miner algorithm, to discover and assess workflow patterns and bottlenecks in the loan application process. Key activities were examined, including W _ Validate Application, W _Call after offers, and W _Handle leads, which exhibited excessive processing times. The analysis revealed significant delays, particularly in the W _ Validate Application step, which took an unprecedented 103.3 years. Redundant activities were also identified, negatively impacting the overall efficiency of the process. The integration of Generative AI facilitated accurate data analysis, but the need for expert interpretation remains critical for actionable insights. To enhance process efficiency, implementing Robotic Process Automation (RPA) and Generative AI in areas with identified delays is recommended. Improving interdepartmental communication and coordination is essential for a streamlined workflow. Additionally, utilizing AI systems for risk detection in complex steps, such as W _Assess Potential Fraud, can significantly improve processing speed and accuracy. This study contributes to the field of process improvement by highlighting the effectiveness of combining Process Mining with Generative AI., offering a systematic approach to identify and address bottlenecks in business processes. The findings provide valuable insights for organizations aiming to enhance operational efficiency.