Generative AI could be the key to speeding up instant payments in the U.S., which has been slower to adopt the technology than
That’s not to say that instant payments aren’t
“If we look at the U.S. to bring it home, there are three major initiatives building up: Same-day ACH, The Clearing House real-time payments and the famous
Banks are already prioritizing instant payments. Last month, BNY partnered with the
That was the culmination of a two-year process, said Carl Slabicki, co-head of global payments for BNY’s Treasury Services.
“Our value prop is really — given our size, scale, experience within this space, the reach of our network and our expertise and payments and kind of being at the bleeding edge of instant payments, both domestically and now internationally — our goal is to help clients send money across the globe instantly,” Slabicki said.
Access to information is key to facilitating instant payments, Capgemini’s Ghanem said. “Instant payments require agility. Regulators are changing the game on a regular basis, so you must be able to adapt.”
Generative AI has multiple use cases in the payments world. In fact, 80% of payment executives said they are using or planning to use generative AI-based tools to support some part of their payments business in the next 24 months, according to the 2024
But generative AI can also be used to speed up instant payments, said Tom Hewson, CEO of RedCompass Labs, a payments modernization technology and consulting firm.
“AI can read every document you’ve ever written in the bank, and it can understand the context of it, and it can compare that to every rule that a regulator has written about how a payment needs to be processed,” Hewson said. “And it can pull that information together and present it back to you in a way that your project team can help you catch up with the rate of change.”
In a separate RedCompass Labs
Generative AI models are relatively simple at their core, Hewson said. “Brute force just says, ‘I’m going to predict the next word. I’ve seen so many trillions of words, I’m going to predict the next one.'”
That works well with modern-day payment structures, which are largely XML- or ISO-based.
“If you look at the payment itself, the ability to read that payment is semantic,” Hewson said. “The tag semantically says, ‘Who’s the sender?'”
Generative AI can also run required checks against the ultimate beneficiary, he said.
“But imagine that every payment has the ability to have an investigation against it. You’re only limited by compute power,” Hewson said. “We moved our payments into human readable code — XML — now the possibilities are limitless.”