This study explores how AI is perceived and operationalized in credit risk assessment within Iranian banking institutions, with a particular focus on the experiences of electronic banking professionals in Tehran. Drawing on grounded theory methodology and semi-structured interviews with 38 practitioners from both public and private banks, the research reveals a complex landscape of technological promise and institutional constraint. Participants emphasized the efficiency, consistency, and expanded analytical reach afforded by AI models, particularly in leveraging alternative data and enhancing fraud detection. However, these benefits are tempered by operational challenges, including fragmented data systems, outdated IT infrastructure, and opaque algorithmic outputs. Ethical and regulatory concerns—especially surrounding algorithmic bias, accountability, and the absence of formal oversight—emerged as significant barriers to responsible deployment. Moreover, organizational resistance, hierarchical decision-making structures, and cultural skepticism toward automation further complicate adoption. The findings suggest strong practitioner support for hybrid decision-making models that integrate AI capabilities with human expertise. This model offers a viable pathway toward responsible innovation, balancing the computational advantages of AI with the contextual judgment and ethical sensitivity of human agents.
