A Comprehensive Review on Financial Fraud Detection and Agentic AI

The exponential growth of digital financial ecosystems has intensified the scale and sophistication of financial fraud. This paper presents a comprehensive survey of fraud detection techniques spanning machine learning (ML), deep learning (DL), and emerging generative AI (GenAI) approaches. Recent advancements (2021–2026) are critically analyzed with respect to accuracy, adaptability, and real-world deployment constraints. The study highlights persistent challenges such as class imbalance, concept drift, and lack of autonomous decision-making. To address these limitations, an Agentic AI-based fraud detection framework is proposed, emphasizing autonomous reasoning, continuous learning, and multi-agent collaboration. The paper also integrates explainable AI (XAI) mechanisms to ensure transparency and regulatory compliance. Comparative analyses and architectural insights provide a foundation for next-generation intelligent fraud detection systems.