Artificial Intelligence Adoption in Banking: A Systematic Review and Integrative Conceptual Framework

This study introduces contradiction zones as a novel analytical prism to explain why artificial intelligence (AI) adoption in banking varies across contexts. Rather than treating inconsistencies as random deviations, contradiction zones incorporate systematic patterns in which constructs such as trust, perceived usefulness, and ease of use yield different results under varying circumstances. Drawing on realist synthesis and configurational approaches , this systematic literature review of 162 peer-reviewed articles published during 2018–2026 follows PRISMA 2020 guidelines and the Theory Context Characteristics Methodology (TCCM) framework to examine the temporal and conceptual evolution of research. The synthesis identifies three structural sources of empirical inconsistencies: unmodeled mechanisms, unspecified contextual conditions, and constructs’ ambiguity. Constructs such as trust are not stable predictors but rather context-dependent variables . influenced by algorithm transparency, fairness, and services’ quality. Finally, the proposed multilevel, mechanism-driven framework explains how antecedents relate to AI adoption through mediating processes across different contextual conditions.