This is the first in a four-part series on disruption in the payments industry.
Terms like "large language models" and "generative AI" were barely part of the vernacular as recently as a year ago, but have become a substantial part of strategies at big payment companies.
"The biggest risk of gen AI is not using gen AI," said Rohit Chauhan, head of AI at Mastercard.
Mastercard and Visa have made investments in gen AI that will drive large-scale global changes in everything from what consumers hear at the call center to what they see on a website to how their transactions are kept safe. The card networks face competition from fintechs that are also bullish on gen AI, making the technology one of the biggest sources of disruption to hit the payments industry in years.
"Gen AI is the transformative technology of our time, and it will have an enormous impact," said Rajet Taneja, president of technology at Visa.
Definitions vary, but gen AI is generally considered to use advanced machine learning and large language models to produce original content or programming, whereas more traditional AI uses machine learning to improve the performance of an existing program over time.
Companies are quickly ramping up investments in gen AI, according to
Firms are expecting operational performance from gen AI, according to Infosys, which means improvements in actual processes as opposed to using gen AI to produce content. And 95% of C-suite executives are in favor of investing in gen AI, suggesting an unusually high level of senior executive buy-in for a new technology, Infosys reports.
For the payments industry, that means work on more tangible projects involving gen AI, and less experimenting, according to Ashvin Parmar, global head of insights and data for financial services at consulting firm
"This year was about internal use and tests because ChatGPT and gen AI are coming in so fast that people are trying to figure out what are the right uses," Parmar said. "Looking ahead, they're going to be thinking about cost containment and more specific results."
In the cards
Both Visa and Mastercard have an active pipeline for gen AI development.
Mastercard just debuted Shopping Muse, a gen AI tool from subsidiary Dynamic Yield that translates general shopping language into specified product recommendations.
Mastercard acquired Dynamic Yield, a company that develops personalized product recommendation engines. Mastercard purchased Dynamic Yield from McDonald's, which acquired Dynamic Yield in 2019 to personalize menus and marketing.
Mastercard's recent investments in gen AI include a $300 million investment in Network, a firm that builds AI to access data from millions of transactions to spot potential fraud.
Shoppers use the tool to enter terms for basic products and receive options to buy based on their payments history, shopping history, demographics and other data. The goal is to enable consumers to enter a term such as "shoes" or "beach clothes" and receive options to buy based on a much deeper profile of that consumer — making the product and incentive offer more tailored, and providing a way to buy that product directly.
"We can use gen AI to add to websites. Instead of going into product menus and dropping down, we can create a place where people can have an interactive conversation and get a clear picture of a product," Chauhan said
New activities at Network International focus on us bringing Mastercard's Brighterion AI technology to support Network's transaction fraud screening and merchant monitoring.
Mastercard also has added gen AI to its global customer care center. The card network will receive about 16 million inbound calls in 2023, with calls increasing at about 20% year over year. Mastercard has developed a proof of concept to use gen AI and large language models to improve responses to inbound queries.
"Eventually this will drive down the cost of customer service down to zero, given how powerful these large language models are," Chauhan said.
Visa has also made several moves in gen AI. Visa has eight patents for gen AI and has published 25 academic papers on the technology. The card network additionally has a pipeline of active products and products in development that use newer forms of AI, including improving speed and accuracy for settlements, managing fraud and driving payments orchestration.
In October, Visa launched a
"The past year has been about building new AI," Taneja said. "Now we're looking to improve speed to market for our larger innovation vision."
Called RTP Prevent, the product produces a risk score for real-time payments, enabling banks to make a call on approval before the payment is executed. The scores range from one for highest-risk payments to 99 for payments with little risk. The product is designed to be used globally.
"Gen AI helps users access data they need from all parts of Visa at any point in time and bring that data together," Taneja said. "This makes the information more valuable and accessible when users need it."
Visa is also using gen AI to develop a payments orchestration product for its
Visa has additionally increased its use of gen AI for internal practices, such as IT projects. Gen AI aids what's called pair programming, which is traditionally defined as one person writing code while another person reviews that code at the same workstation. In this case, a large language model generates different programming options that help the developer produce code.
This process helps build software faster, according to Taneja. "This enables us to do a lot of cybersecurity work, locating and fixing bugs," he said. "It gives you more capacity and increased productivity."
Paying it forward
Block over the past year has introduced nearly a dozen gen AI features covering internal workflows, as well as external uses such as generating images and content for small-business marketing programs.
"ChatGPT can produce more contextualized content," CapGemini's Parmar said. "That is what you're looking for in marketing, service or reporting and analytics."
AI will also embed deeper into processing for international payments, according to Parmar. The
The maturation of gen AI will inform this messaging, if not produce it outright, according to Parmar.
"There will be work around improving producing information in invoices, and how you improve processing," Parmar said "There's going to be a higher speed and higher volume of information."
The next year is expected to include initiatives that embed gen AI more deeply into business trends in the payments industry given the pressure to balance the secular move to faster digital processing and e-commerce with the need to manage IT expenses.
"Most companies are starting pragmatically, doing foundational work either internally or with customer service," said Shannon Johnston, executive vice president and deputy CIO of Global Payments. Johnston will become CIO of the payment technology company early in 2024. "2024 is the year of implementation."
Johnston is monitoring how banks and other companies apply gen AI to embedded payments, or the concept of using payment credentials to access other products.
Gen AI can additionally improve the coding that goes into connecting an enrolled payment account to outside services to improve embedded finance, Johnston said.
Global Payments is also developing new uses for gen AI to contribute to fraud prevention in the next year. Other work includes using the technology to improve payment dispute resolution. There is a case to be made that gen AI can improve the algorithms that are used to manage chargebacks and other disputes, Johnston said. "Merchants don't want to spend time in disputeland."
Global Payments s is looking at gen AI to aid a service that communicates with stakeholders about how technology issues have been resolved.
There are still limits to how payment firms use gen AI, or are experimenting with the technology. While payment companies that offer buy now/pay later lending are aggressively using AI for internal staff work or to improve marketing, decisions on actual lending are off the menu.
"Gen AI is not widely used in credit assessments right now," said Gordon Campbell, co-founder of Rich Data Company, which uses AI to aid lending and credit terms for buy now/pay later loans. AI and gen AI is used similar to payment fraud prevention for BNPL lending, producing a score that can guide decisions, but not make them, Campbell said.
Writing for
"With credit decisions you have to be careful about how you explain the outcome," Campbell said.
