AI’s Time Has Come

Machine learning has long been a tool used by payments companies. Now, artificial intelligence is poised to offer even more utility. Safeguards will be needed.

Artificial intelligence is on the cusp of changing how payments companies interact with each other and extract insights from the billions of units of data they collectively hold and gather. All in all, AI is about to make some aspects of payments processing easier.

But adoption of the technology is not without costs. How that will play out, which factors will be most influential, and which basic choices payments companies make regarding how they employ AI will determine their success.

As defined by IBM Corp., artificial intelligence is a field that combines computer science and robust datasets to enable problem solving. It is not a new endeavor for the payments industry. Machine learning, a branch of AI that uses data and algorithms to imitate the way humans learn, has been in use in the payments industry for years, again according to IBM.

“AI models communicate and appear to reason like human beings,” says Paul Harrald, chief financial officer at Curve Ltd., a London-based fintech, “whereas machine-learning models are numerical computations. In effect, AI models influence human observers in a way that ML models do not—they appear to communicate and reason to be creative.” AI today is sometimes labeled generative AI, or genAI for short, because it can generate content.

“AI will undoubtedly expand the scope of opportunities for banks, fintech companies, payment services providers, and other entities in the financial sector,” says Tue To, head of advanced payments and fintech for North America at Edgar, Dunn & Co., a San Francisco-based payments consultancy.

“We are currently witnessing only the initial days of generative AI,” says Luis Silva, chief executive and founder of CloudWalk Inc., a Brazil-based payment provider planning a U.S. expansion for this year. “Existing forms of AI are propelling innovation in payments, and increased customer satisfaction.”

How this may play out, and specifically how payments companies may incorporate AI into their operations beyond any direct payments application, could add value beyond the transaction itself.

The Ideal Problem

“There are a number of promising applications for [generative AI] in the payments industry,” says Shannon Johnston, chief information officer and senior executive vice president at Atlanta-based Global Payments Inc. “I would put them in five main categories: managing fraud and risk; enhancing operational customer support; delivering friction-free payments; enhancing operational efficiencies; and enabling new payment products.”

Johnston says managing fraud and risk with AI assistance has emerged as an obvious application. “Currently, fraud-detection systems have too few ‘genuine’ or strong fraud cases to analyze and learn from,” she says. “With genAI, you could produce synthetic examples of fraud based on the patterns established by actual cases. These sequences would, in turn, help improve fraud-detecting systems.”

Other observers similarly view fraud as an ideal problem for AI applications to work on.

“Payments fraud has gotten significantly worse in recent years,” says Tony DeSanctis, senior director at Cornerstone Advisors, a Scottsdale, Ariz.-based firm. AI could improve the ability to identify fraud and risky behavior, he says. Like many current tools, AI could be used to reduce fraud to zero. But that would risk impairing a consumer’s ability to make a payment and a merchant’s ability to accept one.

As an example, AI could help with the auto-decision process that occurs when a consumer who reliably purchases a cup of coffee once a day suddenly makes a purchase of six laptops, DeSanctis says.

And while machine learning has been used in helping with fraud mitigation, AI could advance these efforts, says Malcolm DeMayo, vice president of financial services at Nvidia Corp., a Santa Clara, Calif.-based computing-technology provider. There may be an envelope of only 1,500 milliseconds—from tap or dip to a go-no-go decision from the issuer—to determine whether the transaction is valid or potentially fraudulent, he says.

“Being able to identify fraudulent activity in that 1,500-millisecond envelope is really super important,” DeMayo says. AI can help improve the accuracy of these decisions, he says.

AI-enabled companies will be able to do more, DeMayo says. “Technology has always enabled us to do more. AI is allowing us to do more better.”

That’s how fintech One Inc is viewing AI’s potential for its operation. Based in Folsom, Calif., One Inc focuses on payment services for the insurance industry.

AI could produce multiple benefits for the payments industry, says Ian Drysdale, One Inc’s chief executive, among them more secure transactions, optimized processing that lowers costs, and enhanced experiences.

“With the ability to understand and adapt to emerging fraud patterns, generative AI can continuously evolve security measures. This adaptability is crucial in staying ahead of ever-evolving threats in the landscape of payment security,” Drysdale says.

“Generative AI can also assist in making data-driven decisions by analyzing vast amounts of transaction and customer data,” he adds. “This efficiency can lead to operational and strategic improvements, benefiting both payment providers and consumers.”

As DeSanctis says, AI employed to aid messaging could avoid the problem of consumers receiving multiple messages that repeat previous ones or don’t apply to that consumer.

As an example, DeSanctis says he recently opened an account with an online financial-services firm. He then received four emails a day at least three days a week from that firm. “I got emails about student-loan refinancing,” he says. “I haven’t had a student loan in years.”

AI could also help improve the customer focus and make marketing messages more personalized. “For end customers, if the AI application is more customer-focused, we could see less friction and more fine-tuned step-ups,” says Ananth Gundabattula, cofounder and senior architect of AI, data, and privacy at Darwinium, a fraud-prevention platform for payments providers and fintechs.

“There could be financial gains for payments companies as a result of leveraging GenAI to make better-informed recommendations of exchange rates, dynamic pricing [and so on],” he says.

‘You Have to Be Smarter’

AI has a lot of potential benefits, but there is a cost. First, across many cultures, there is some broad distrust of artificial intelligence that comes without human control. The European Parliament has proposed the AI Act, which would require the use of AI to be declared when users interact with applications that use it.

And, as DeMayo says, AI is a tool that anyone can use, and that includes bad actors. “You know some bad people are going to get their hands on it,” he says. “The only way to fight that is to make sure we stay ahead.”

There also are concerns about a lack of transparency regarding AI’s decision-making protocols. “While there are lots of benefits with the usage of AI, it also raises valid concerns regarding lack of transparency … potential discrimination or bias, data privacy, and the absence of human empathy in interactions,” says Edgar, Dunn’s To.

“Balancing automation while preserving a human touch and ensuring robust data privacy and security, coupled with ethical considerations, is crucial for long-term success in business,” To argues.

Johnston at Global Payments suggests financial, environmental, and governance issues may impede AI growth. “In a conservative environment in terms of macroeconomic headwinds, brands will be careful about spending substantial resources on AI unless it truly has a counterbalance of cost savings. Many CIOs will be asking, ‘How do I make this technology affordable?’” Johnston says.

The energy to power the considerable computing needs of AI also could affect a company’s carbon footprint and its energy use, she says. A lack of a formalized policy or strategy could be an impediment, too. Such a policy “includes providing a forum for people across an enterprise to talk to each other, so they are sharing their knowledge around best practices, successes, and failures,” Johnston suggests.

Another issue may be the perception that AI will replace human workers, especially those in programming positions.

“AI will not replace people,” DeSanctis declares. “People who know how to use AI will replace people who don’t know how to use AI. You have to be smarter.”