AI-Enabled Automation of High-Risk Decision-Making Processes
Abstract
Intelligent Process Automation (IPA) offers strong potential for automating business processes through the application of Artificial Intelligence (AI). However, applying AI to high-risk decision-making processes remains challenging due to regulatory requirements as well as technological and organizational factors. In particular, the EU AI Act imposes strict requirements on human oversight, thereby limiting the potential for automation in high-risk environments. In this research-in-progress paper, we address this gap by conducting an Echeloned Design Science Research (eDSR) project in collaboration with a financial institution to design an AI system for automating the loan approval process. Our problem analysis identifies several challenges, including time-critical and complex decision-making as well as regulatory requirements such as mandatory human oversight, decision justification, accountability, and data privacy. Based on this, we derive validated design objectives and requirements, and we position risk mitigation as a core design challenge. We also provide an outlook for the subsequent echelon types to continue this research project.
Keywords
Intelligent Process Automation; Design Science Research; High-Risk Decision-Making; Artificial Intelligence; EU AI Act