Firms Add Humans Alongside Fraud AI
Despite soaring AI adoption in fraud and AML — with 98% of firms now using it — 94% still plan to grow their human teams. The data suggests a strong continuing need for expert oversight and explainable models, rather than a full replacement of personnel with black-box automation.
The push for "explainable AI" (XAI) is a direct response to the "black box" problem, where even developers can't fully dissect how a model reaches a decision. Regulators, particularly under frameworks like the EU's GDPR, are reinforcing the "right to explanation," compelling firms to justify automated decisions, such as a flagged transaction, to customers and authorities. This is driving investment in XAI techniques like LIME and SHAP, which help analysts understand the specific factors leading to a fraud classification. The urgency is compounded by fraudsters leveraging AI for their own ends. AI-driven fraud now accounts for 42.5% of all detected attempts in the financial sector, with tactics including deepfakes to fool KYC systems and AI-written phishing emails. This escalates the need for human-led teams that can analyze and respond to novel attack patterns that AI, trained on historical data, might miss. This dual human-machine approach is also shaping underwriting, where AI models can process vast alternative datasets—from utility bills to digital footprints—to assess "credit invisibles." This can slash processing times by up to 80% and improve accuracy, but human oversight remains essential for complex, regulated decisions and to manage risks of algorithmic bias. On the infrastructure side, the growth of real-time payment networks creates new fraud vectors. The RTP® network saw its payment value jump 94% to $246 billion in 2024, with daily transactions now exceeding 2 million. Concurrently, the FedNow service saw its quarterly volume grow 62% in Q2 2025, highlighting the rapid market shift toward instant settlement. Digital identity solutions are becoming critical to securing these new rails. By using biometrics, multi-factor authentication, and tokenization, firms can enhance security for instant payments and streamline compliance with KYC and AML regulations. This creates a verifiable layer of trust necessary for a high-speed, automated payments ecosystem. Institutional adoption of stablecoins for cross-border payments and treasury management is also accelerating, with projections suggesting they could handle $2.1 to $4.2 trillion of cross-border payments by 2030. Major players are exploring stablecoins for faster settlement and improved liquidity, adding another complex, high-value transaction type that requires sophisticated, human-overseen fraud and AML monitoring.
