Responsible AI Adoption in Nepalese Banking: A Structured Review

Abstract

Artificial intelligence (AI) is becoming more common in Nepalese banking, where it supports customer service, fraud monitoring, credit assessment, personalized services, investment analysis, and internal decisions. Yet the evidence is scattered across journal articles, theses, policy documents, institutional reports, and review papers. This structured review synthesizes 20 unique sources from an initial register of 31 records. After three duplicate, parallel-language, or incomplete records were removed or consolidated, 28 records were screened and eight were excluded for limited relevance or insufficient evidence. The final set included Nepal-specific studies, government and regulatory documents, comparative research, and contextual sources. The evidence indicates that AI can improve decision speed, credit scoring, service automation, personalization, and investment analysis, although the reported benefits vary by application and study quality. It also identifies persistent weaknesses in model transparency, validation, vendor oversight, data privacy, and responsibility for harmful decisions. Employees are often comfortable with ordinary digital tools but less able to explain AI systems or challenge their outputs. Customer satisfaction and financial inclusion depend not only on convenience but also on security, understandable explanations, digital skills, accessible design, and continued human support. Overall, AI adoption is advancing faster than the arrangements needed to govern its risks. The article therefore proposes a practical framework based on risk classification, data and model checks, clear human authority, staff training, customer protection, and continuous monitoring. AI can strengthen Nepalese banking only when technical development and responsible governance progress together.