Factors Influencing the Effectiveness of AI-Based Fraud Detection

Abstract The increasing sophistication and volume of financial fraud in digital banking systems have compelled organizations to adopt artificial intelligence (AI) as a core detection mechanism. This study investigated the influence of three critical AI system attributes—perceived AI accuracy, AI speed, and AI reliability—on the perceived effectiveness of AI-based fraud detection among 300 professionals in the banking and financial services sector. Grounded in the DeLone and McLean Information Systems Success Model, a quantitative cross-sectional survey design was employed, and multiple linear regression analysis was conducted. The regression model explained 85.5% of the variance in fraud detection effectiveness (R² = .855, F(3, 296) = 582, p < .001). All three predictors were statistically significant: AI Speed exhibited the strongest standardized effect (β = .350, p < .001), followed by AI Reliability (β = .319, p < .001) and AI Accuracy (β = .298, p < .001). These findings underscore the importance of optimizing not only the accuracy but also the processing speed and reliability of AI fraud detection systems.