A Comprehensive Review of Financial Fraud Detection Techniques and the Emerging Role of Adaptive Agentic AI

Abstract

 The rapid growth of digital banking, online payments, and real-time financial services has significantly increased financial fraud, creating major economic and operational challenges for banks. To mitigate these risks, a wide range of artificial intelligence (AI) techniques—including rule-based systems, machine learning, deep learning, graph-based, and agent-based models—have been widely adopted. However, many existing approaches remain limited by static behavior, weak adaptability to concept drift, high false-positive rates, and limited transparency in decision-making. This review analyzes peer-reviewed studies published between 2015 and 2025, covering traditional and AI-driven fraud detection methods in banking and finance. We examine key research trends, highlighting the shift toward sequential modeling, graph-based reasoning, and explainable AI, while identifying challenges that affect real-world deployment. The paper further discusses Agentic AI as an emerging paradigm that supports autonomous decision-making, continual learning, and coordinated multi-agent reasoning. Finally, open research challenges and future directions are outlined to support the development of robust, adaptive, and trustworthy fraud detection systems..