Banks Deploy Agentic AI to Trace Stolen Payments

Nasdaq Verafin announced on June 10, 2026 the expansion of its Agentic AI Workforce with two new role-based agents: an Agentic Fraud Analyst, which will initially triage unusual ACH activity, and an Agentic AML Analyst, which will focus on cash-structuring alerts (cases where criminals break up large sums to avoid regulatory reporting thresholds) before expanding to flow-of-funds analysis and unusual international transactions. Both agents reach general availability in the third quarter of 2026. Verafin says more than 650 financial institutions have already adopted its platform, which runs on a consortium data network spanning more than 2,800 institutions, letting the system flag counterparty fraud risk across banks rather than within just one. The company reports its existing agents have already cut workloads meaningfully: the Agentic Sanctions Analyst reduced alert review by up to 90%, and the Agentic EDD Analyst cut enhanced due diligence review time by up to 50%.

PYMNTS frames the announcement against a worsening fraud picture. PYMNTS Intelligence found 40% of financial institutions lost more money to fraud last year and 38% saw higher fraud volumes; scams now account for 23% of fraudulent transactions after a 56% year-over-year rise, with the dollar value lost to scams up 121%. In the U.K., authorized push payment fraud, where the victim authorizes a payment to a scammer with valid credentials, rose 19% to 576.4 million pounds (about $774 million) last year, with 66% of cases starting on online platforms. Because real-time rails clear funds irreversibly, banks increasingly compete on how fast they can trace money after it moves rather than on blocking it beforehand. Similar tools are emerging elsewhere: India's Reserve Bank Innovation Hub operates MuleHunter.AI across 26 banks, detecting roughly 20,000 mule accounts a month; India's Cyber Crime Coordination Centre had identified 2.65 million first-layer mule accounts as of December 31, tied to nearly 200 billion rupees (about $2.4 billion) in alleged theft, of which about 82 billion rupees (roughly $980 million) has been recovered. Separately, JPMorgan Chase and ACI Worldwide announced a partnership embedding JPMorgan's Kinexys Liink account verification directly into ACI's fraud platform, applying controls before funds leave an account, a complementary, pre-clear approach.

Agentic systems built for post-clear tracing typically chain queries across internal ledgers, external account-enrichment APIs, transaction graphs, and identity signals to reconstruct money flows, which cuts repetitive manual lookups but concentrates new risk around model explainability, the provenance of enriched data, and false leads that still require human validation. Deployments that follow this pattern typically need tighter role-based access controls, immutable logging for regulatory audits, and integration with SAR/STR reporting workflows; observability into agent decision paths and conservative escalation policies are the common mitigations. Note that Verafin's two new agents are not yet generally available, they ship in Q3 2026, so the concrete efficiency numbers currently public (90% and 50% reductions) describe its existing Sanctions and EDD agents, not the new Fraud and AML analysts.

Whether Verafin or its bank customers publish measured false-positive and false-negative rates once the new agents reach general availability in Q3 2026; whether cross-institution tracing prompts new data-sharing arrangements or regulatory guidance; and whether pre-clear approaches like the JPMorgan-ACI partnership reduce how often post-clear tracing tools are actually needed.