The opinions expressed here are those of the authors. They do not necessarily reflect the views or positions of UK Finance or its members.

That era is ending. Agentic AI is now embedded in the systems banks rely on, able to take action without waiting for a human. 

It’s a powerful shift. Done well, it reduces manual work, scales decision-making, and reacts in milliseconds. Done badly, it adds confusion and erodes trust. Meanwhile, fraudsters are using generative AI to launch attacks at scale, adapt to defences, and hit harder. AI is already part of the equation. The question is whether it’s helping you or them. 

Most fraud doesn’t break the system, it uses it perfectly 

Authentication passes. Transactions are approved. Everything looks fine until the money moves. That’s the problem. 

Traditional defences focus on the perimeter: firewalls, MFA, and access controls. But those protections fade once a session is authenticated. Much of today’s fraud happens inside the session, long before payment. 

Agentic AI introduces a new layer of defence. It coordinates signals across systems, adapts to threats in real time, and streamlines analyst workflows. Rather than replacing core detection methods like in-session monitoring capture, agentic AI acts as a force multiplier, helping teams manage scale, complexity, and decision fatigue with greater precision.

 APP scams prove the system isn’t broken, just blind 

APP fraud is rising fast. To many, APP scams seem unsolvable. But they’re not. These scams don’t usually start with a phone call. They begin with reconnaissance, setup, and session manipulation – all of which leave detectable signals. 

With high-quality telemetry and well-placed models, those signals become actionable. Real-time session intelligence lets banks spot fraud before the payment is made, giving them time to intervene. 

Scams are scaling from one-to-one to one-to-many, adapting tone, timing, and tactics in real-time. Defenders need automation that understands context, prioritises the right actions, and supports human decision-making. That’s where agentic AI comes in. 

Seeing fraud before it happens 

Fraud starts quietly, with scammers gathering information and making subtle moves while users are logged in. That’s why it’s critical to monitor what’s happening during the session, not just at the point of payment.

Real-time session intelligence breaks down user activity into smaller parts to flag signs like remote access or unusual behaviour early enough to act. 

This isn’t about replacing people. It’s about helping fraud teams do more, faster. Generative AI co-pilots can summarise key information and suggest next steps, leaving analysts to focus on complex cases. 

Rather than relying on one system to do it all, this approach layers multiple signals to support decisions mid-session, cutting through the noise to highlight the moments that matter/0

Helping fraud teams move faster, with less noise 

Good teams know what to do. The challenge is scale, case volumes, edge cases, and time pressure. When overloaded, key signals get missed. 

Agentic AI helps by knowing when to step in and when to step back. Unlike basic automation, which speeds up routine tasks, agentic AI is goal-driven. It understands intent, reacts in real-time, and keeps a clear audit trail. 

More automation isn’t always better. If misapplied, it can misjudge intent, act opaquely, and erode trust. The answer isn’t less automation, it’s smarter automation. Agentic systems must be built with guardrails, transparent logs, and escalation protocols to return control when uncertainty is high. That’s how you scale trust, not just output. 

Fraudsters have embraced automation

They’re moving fast, adapting quickly, and operating at scale. It’s time we did the same. To see how early fraud detection can support your bank’s efforts, visit cleafy.com