Agentic commerce sounds futuristic, until it starts hitting your auth flows, your dispute queues, and your monitoring pipelines.

In this episode of Fraud Forward, host Hailey Windham sits down with Chen Zamir, Head of Fraud Strategy at Sardine and Founder of Native Risk, to unpack what happens when AI agents don’t just recommend, they act. They browse, click, checkout, retry, and optimize at machine speed, sometimes with no human in the loop.

Hailey and Chen break down the two flavors of agentic commerce (API/MCP-based vs. in-browser agents), then get direct about the fraud pressure that follows any new payment-adjacent product. They dig into the first typologies likely to spike, why “secure protocols” won’t solve the real problems, and the OTP timing trap that makes step-up friction feel irrelevant when the customer is asleep, offline, or busy on purpose.

They also cover the downstream damage teams aren’t modeling yet: rising abandonment, risk scores inflated by failed challenges, messier proof of intent, and a stack that struggles to separate agentic flows from everything else.

The core takeaway is simple: fraud teams need to identify, route, and manage agent-driven transactions as a distinct channel before the ecosystem forces the issue.

Guest lineup:

1. Chen Zamir: Head of Fraud Strategy at Sardine, founder of NativeRisk
2. Hailey Windham: Host of Fraud Forward and Community Banking Lead at Sardine