Artificial Intelligence and Fraud Detection: An Empirical Study of Selected Deposit Money Banks in Nigeria

Abstract

This research investigated how Computer Vision (CV) and Robotic Process Automation (RPA) influence fraud detection capabilities in five quoted Deposit Money Banks (DMBs) in Nigeria between 2020 and 2026. The study was anchored on the Fraud Triangle Theory and the Technology Acceptance Model (TAM) and adopted a descriptive survey and correlational design. Primary data were gathered from 88 personnel drawn from Information Technology, Risk Management and Internal Control units, while secondary data on fraud losses were extracted from the publications of the Central Bank of Nigeria (CBN), Nigeria Deposit Insurance Corporation (NDIC) and Nigeria Inter-Bank Settlement System (NIBSS). Data analysis involved descriptive statistics and panel random-effects regression. Results indicated substantial adoption of CV (Grand Mean = 3.81) and moderate adoption of RPA (Grand Mean = 3.51). CV exerts a significant positive influence on fraud detection rate (β = 0.341, p < 0.01), whereas RPA significantly lowers fraud response time (β = -0.298, p < 0.01). Both CV (β = -0.213, p < 0.01) and RPA (β = -0.189, p < 0.05) significantly curtail fraud losses, explaining 63.1% of the variation (R² = 0.631). Major constraints include low-quality document images and regulatory restrictions on automated decision-making. The paper recommends expanding CV to cheque and mandate verification, deploying RPA for real-time preventive controls, and establishing a CBN-supervised sandbox for automated fraud actions.