AI beats humans at transaction monitoring, Revolut US CEO says

Revolut's U.S. CEO Cetin Duransoy and Elana Schor, senior Washington editor at Semafor.
Revolut's U.S. CEO Cetin Duransoy and Elana Schor, senior Washington editor at Semafor.

AI has quickly moved from add-on to core tooling, Revolut's U.S. CEO Cetin Duransoy said at Semafor's Banking on the Future Forum in Washington, D.C. on Wednesday.

Processing Content

He argued that AI is now essential infrastructure that helps the company track and adapt to regulations across 39 countries. The fintech has also deployed agentic AI for know-your-customer and anti-money laundering use cases.

For transaction monitoring, AI performs "statistically significantly better than human reviews of the transactions," he said. "We are now complementing our human reviewers [who are] looking into much more complex cases." 

Every employee uses AI to improve productivity, and AI has also been powering the company's customer service interactions through tailored responses to queries based on user data.

"Whenever you ask a question, you get an LLM-based answer based on actual account information," he noted. "Obviously, hallucination is a big problem that needs to be controlled incredibly well, and we train accordingly … especially in the customer service area very strictly."

Other fintech executives at the event were also touting AI's benefits. 

Harsh Sinha, U.S. president and chief technology officer at Wise, said the payments company deployed an assistant that handles roughly 50% of customer queries via LLMs. 

"What you want is the tuning such that the quality of the responses are really good…so that's why we've been careful in how we roll this out," he said, noting that in many cases AI responses are better than what humans could provide.

Fighting AI-powered fraud

Duransoy warned that AI is also being turned against consumers, with fraudsters building their own agentic tools capable of generating tailored personas to target victims without any human involvement.

The current system puts the burden entirely on financial institutions to catch and cover losses. He said he supported the Scam Act, a bill introduced in February that would prohibit online platforms from displaying fraudulent or deceptive paid advertisements and require them to verify advertiser identities. It would strip platforms of the protections of Section 230, the legal shield that has protected social media platforms from liability for third-party content.

"A material portion of the revenue of some of the social media companies is coming from fraudulent posts, and right now incentives are not aligned," he said. "Social engineering became so much easier in the last 15 years … now it has gone to a next level with agentic AI, because you don't even need humans to replicate another human to get into someone's mind."

He acknowledged the bill faces a steep lobbying battle, saying he had heard of large sums being spent by social media companies to fight it, though he declined to name them. With as much as 15% of revenue potentially at stake, he said, the math on lobbying spend makes sense for these platforms.

Duransoy declined to weigh in on the Trump administration's executive order on fintech, unveiled the day before the event, but suggested more broadly that the industry welcomes the administration's openness to innovation.

U.S. expansion plans

Revolut is awaiting a U.S. bank charter, working through an approval sequence with the OCC, the FDIC and the Federal Reserve. The application is backed by $6 billion in revenue, $2.3 billion in net profit and a commitment to invest $500 million over three years in capital, marketing and people.

"There is a very speedy approach from regulators to get this done," he said, noting the OCC has a stated 120-day KPI for preliminary approvals.

Once the bank launches, Duransoy said Revolut plans to grow its U.S. consumer offerings, including its flagship foreign exchange product.

"A good product sells itself," he said. "Consumers are going to see it both through word of mouth as well as by just testing it on their own."


For reprint and licensing requests for this article, click here.
Artificial intelligence Bank technology
MORE FROM AMERICAN BANKER
Load More