JPMorgan Payments is blending generative AI with machine learning to improve fraud prevention.
“I often joke, we can catch 100% of the fraud by stopping every payment and investigating it, but that would result in a horrible customer experience,” Greg Hodges, head of trust and safety at JPMorgan Payments, told FinAi News. “But we have to find a balance between protecting the customer without generating a lot of false positives that induces friction.”

The $3.8 trillion bank manages more than $12 trillion a day, Hodges said, adding that gen AI allows it to build on existing machine learning models to reduce the fraud loss ratio.
Payment life cycle
JPMorgan Payments divides fraud mitigation systems into segments:
- Before payment initiation. JPMorgan’s strategy teams have the ability to evaluate the impact of fraud controls to maximize fraud capture while minimizing customer impact, he said.
Bankers can input rules in natural language and create layers of fraud detection that used to require deep technical knowledge, Hodges said. That allows the bank to go from ideation or analysis to delivery and implementation of a rule significantly faster.
“In the past, you would have to go through the entire software development life cycle to get [a rule] implemented,” Hodges said. “Now you can type in what kind of a rule you want to implement, you analyze it, you understand it, you pressure test it, you deliver it to the production.”
- During payment. Machine learning has been used to detect anomalies associated with certain transactions, but gen AI is able to provide those risk scores quicker while reducing false positives, he said. “Graph technologies allow us to gain better insights into relationships of parties or organizations than has been previously available with traditional machine learning techniques.”
- Post-payment. If fraud is found, agentic AI tools can help clean it up by handling consumer complaints and collecting data points of the events, Hodges said. “We’ve had some good success in starting to push the envelope there from an automation standpoint.”
While JPMorgan Payments didn’t disclose efficiency gained by blending AI and gen AI for fraud detection, Hodges said certain parts of the process have seen “double-digit percentage increases.”
Register here for the inaugural FinAi Banking Summit, taking place March 2-3 in Denver. View the full event agenda here.






