As scams intensify, agents confront and fight them
A consumer receives a text message that appears to be from her bank. The tone was urgent. This link will open a compelling landing page. Enter your credentials. After a few moments, the transfer request appears in what appears to be a trusted session.
Across town, the chief financial officer of a mid-sized company was reviewing invoices from a long-term supplier. This amount is consistent with the previous payment. Bank details are new and come with regular explanations. The transaction has been approved. A few days later, the supplier called. The funds never arrived.
These situations reflect a common weakness in modern fraud prevention. The transaction has been authorized. The voucher is valid. Taken in isolation, this behavior seems reasonable. Deception occurs at the level of human judgment, not system destruction.
PYMNTS Intelligence data underscores how quickly this model is expanding. Recently, fraud accounted for 23% of fraudulent transactions reported by financial institutions, an increase of 56% year over year. Even more telling is the 121% increase in the percentage of dollars lost to fraud.
Why fraud detection struggles
Fraud systems have traditionally been built around discrete checkpoints, such as verifying logins and scoring transactions. Payment is approved or declined. If fraud later occurs, disputes and reimbursements can ensue.
This structure works well when fraud stems from stolen credentials, compromised cards, or unusual spending patterns. Scam-driven fraud presents different challenges because the payment “intent” appears to be legitimate. Risk signals are often generated during interactions.
AD: SCROLL TO CONTINUE
The real-time payments track adds to this pressure as settlement windows are compressed. Reversal becomes more difficult.
Financial institutions are already feeling the pinch. PYMNTS Intelligence found that 40% of financial institutions (FIs) lost more money to fraud, while 38% experienced higher fraud volumes.
How agents change the control model
Agent artificial intelligence (AI) agents introduce different operational postures in combating fraud. Rather than assessing risk at a fixed point, the system continuously observes and evaluates throughout the transaction lifecycle. In effect, agents can change when and how interventions occur.
As PYMNTS CEO Karen Webster noted, speed and data are critical in this era, with last month's data showing that 41% of consumers use dedicated AI platforms to discover products, and agent-powered shopping could equate to more than $5 trillion in spending activity. “Instead of prioritizing artificial intelligence over old habits, they're closing the door and leaving them behind,” Webster said of these consumers.
Agents can evaluate behavioral patterns as Customer browses payment session. As signals evolve, data and patterns can correlate device properties, session dynamics, historical activity and contextual anomalies. As interactions unfold, decisions are refined. Best of all, the intervention itself does not require disruption, so parties on both sides of the transaction face no problems.
Business without obvious friction
We'd love to be your go-to news source.
Please add us to your preferred sources list so that our news, data and interviews appear among your sources. Thanks!
A common concern surrounding advanced fraud controls is the risk of a slowdown in legitimate activity. In practice, agency systems can operate largely outside the customer's awareness.
The risk assessment is carried out simultaneously with the payment process. Authentication strength can be adjusted using credentials already in the session. Contextual confirmations appear as routine security checks rather than alerts.
This feature resolves a long-standing tension in fraud management. Historically, greater control has translated into greater friction. Agent systems enable more selective responses.
Why do organizations care?
The industry's investment patterns reflect these pressures. PYMNTS Intelligence reports that 26% of financial institutions have added behavioral analytics capabilities in the past year, while 76% are deploying or planning new fraud technologies. Confidence in faster payments is also growing. 98% of financial institutions say they believe they can provide a faster payment experience without compromising security.
External forecasts heighten the urgency. Deloitte expects authorized push payment fraud to grow significantly as adoption of instant payments expands.
Loss prevention strategies increasingly require systems that can dynamically assess intent, context, and behavior.
Agency business case
From an economic perspective, early intervention reduces the severity of losses
Proxies are also aligned with the business interface trajectory. As digital experiences move toward embedded finance and agent-driven interactions, continuous decision-making will become a structural requirement.
Currently, adoption challenges remain significant. Cost pressure continues to dominate investment decisions. PYMNTS Intelligence data shows that 83% of financial institutions cite cost as a limiting factor in fraud prevention upgrades. Governance needs are also expanding. Continuous decision-making introduces model risk considerations, auditability requirements, and liability issues. Over the longer term, as risk assessments become ongoing, agents reflect broader shifts in defense strategies.
