Riskified Adds Identity Risk Intelligence to Zendesk Agents

What happened

Riskified has integrated its identity risk intelligence into Zendesk, giving retailers using the platform's Industry Agents access to fraud and abuse signals drawn from its database of orders, claims and chargebacks. The integration is designed to help customer service teams distinguish genuine returns and claims from fraudulent ones without slowing down or degrading the experience for legitimate shoppers.

In practice, this means Zendesk agents handling return or refund requests can now see risk scoring pulled from Riskified's transaction history directly within their existing workflow, rather than escalating suspicious cases to a separate fraud team or applying blanket friction to all customers.

Why it matters

Returns and refund fraud sit at an uncomfortable intersection for retailers: tightening verification to catch bad actors risks alienating the vast majority of honest customers with extra steps, delays or denied claims. Embedding risk intelligence directly into the support agent's interface is a step toward resolving that trade-off computationally rather than through blunt policy — agents get a real-time signal that lets them approve legitimate claims quickly while flagging likely abuse for further scrutiny.

For CX and digital transformation leaders, the development is another example of AI-driven risk data moving out of specialist fraud-prevention tools and into everyday service platforms, effectively turning frontline support agents into a first line of fraud defence rather than a bottleneck for it.

The Renascence take

The interesting part of this story isn't the fraud-detection technology itself — it's where the intelligence now sits: inside the same screen the agent already uses to talk to the customer.

Most retailers still treat fraud prevention and customer service as separate disciplines with separate owners, separate tools and often conflicting incentives — one team is rewarded for speed and satisfaction, the other for loss prevention. Pushing risk signals into the agent's existing workflow, rather than a side system they have to consult, is the real design principle here: friction should be applied selectively and invisibly to the few, not uniformly to the many. Operators evaluating similar integrations should resist the temptation to let risk scores silently drive automatic denials; the value is in giving human agents better judgment, not removing it.