Financial crime is now just as much a treasury challenge as it is a security one. As threats become more complex, organizations are confronting a new reality: the greatest risk is no longer confined to external breaches alone. It often emerges at the moment of payment initiation, when a legitimate user authorizes a transaction under false pretenses.

This dynamic is not new. For more than a decade, business email compromise (BEC) and impersonation-based fraud have been among the most damaging threats facing organizations, exploiting trust, urgency, and gaps in payment verification rather than technical system failures. What has changed is not the core tactics, but the tools used to execute them. Artificial intelligence (AI) is accelerating this shift.

In the wrong hands, AI enables more sophisticated impersonation and targeted fraud at the point of payment. In the right hands, it equips teams to identify threats earlier and respond with greater precision. In this environment, treasury management is no longer just about moving money; it serves as a critical gatekeeper for risk, governing payment decisions across every touchpoint before funds leave the organization.

AI-related incidents cost victims more than $893 million in 2025 alone.^[1] BEC has become the leading avenue for corporate fraud. In controlled studies, AI-assisted phishing campaigns have shown success rates as high as 81%. Without AI, generic phishing emails typically see engagement rates closer to 19% to 28%.^[3]

Modern attackers now count on employees to act on seemingly routine requests, initiating fraudulent transactions with valid credentials that are far harder for treasury teams to detect. Impersonation remains the dominant driver: most email-based fraud attempts involve criminals posing as third parties or vendors requesting changes to bank account or payment instructions, and a majority of organizations report imposters posing as senior executives to redirect funds.

These capabilities also extend to how organizations manage incoming communication. In controlled testing environments, AI-enabled LLMs have detected malicious intent in phishing emails with up to 75% accuracy, and can recommend actions such as verifying requests through official channels.³

More than a technology upgrade, the shift to proactive intelligence is a necessary evolution in how treasury teams approach risk. Risk mitigation needs to be dynamic, adaptive and forward-looking.

Once these capabilities are embedded within treasury operations rather than layered on top of them, systems can interpret signals consistently and at scale.

Electronic payments provide enriched, standardized data fields through formats such as ISO 20022. Structured data allows systems to interpret key transaction details, such as beneficiary information and payment context, with far greater precision than legacy free-form inputs. It establishes patterns of normal transaction behavior, making it easier to identify anomalies before funds move.

“We get a lot of data intelligence from digital payment transactions,” explains Josh Del Valle, PNC’s Head of Enterprise Fraud. “Initiation is where the risk is, but once that happens, the flow to the beneficiary is controlled systematically.”

“Now, think about a paper check,” Del Valle continues. “After it’s issued, there’s time to alter the check and create copies before the beneficiary receives it, there’s so much risk in that process.”

Compared to electronic payments, paper checks introduce additional points of vulnerability. They offer limited visibility into the transaction and can be tampered with or redirected without immediate detection. AI can only protect what it can see, and paper checks hide too much. Treasury teams moving from paper to electronic payments cut the surface area for fraud and give banks more context to flag unusual activity.