An Exploratory Case Study On The Use of Enterprise Modeling For Ai-Based Fraud Detection

The rise of digitalization has increased fraud phenomena in the banking sector, often exposing the limitations of traditional rule-based approaches. New approaches based on Artificial Intelligence are on the rise, yet, their adoption and implementation pose significant challenges for organizations. Enterprise Modeling can facilitate and support these activities. This paper employs a case study in the banking sector of Sri Lanka to explore and demonstrate this facilitation. Fourteen semi-structured interviews with domain experts have been used to collect data, which were subjected to thematic analysis. The results show that the current approach bears a number of weaknesses, such as a lack of flexibility in the rule-based processes, limitations in data access, and room for improvements in cross-departmental collaboration. Additionally, Enterprise Modeling can support defining roles and responsibilities, improve the auditability of decisions made by Artificial Intelligence, and structure fraud detection procedures. These findings have been used to develop and propose a UML model that extends the rule-based approach with Artificial Intelligence. The model is proposed as a blueprint that could support adoption and implementation. The results suggest that using Enterprise Modeling to guide the implementation of Artificial Intelligence can have a positive impact.