Citizens Financial Group has explored over 90 different use cases for generative artificial intelligence. To date, only two have been put into production.
“We're really committed to making sure we have the right guardrails and security in place for common use cases,” said Michael Rutledge, chief information officer at Citizens Financial.
The Rhode Island-based bank said it has taken a cautious approach to generative AI, including creating a steering committee to keep employees from independently developing their own projects.Two generative AI use cases Citizens Financial plans to roll out this year include an enterprise search tool to help customer service representatives answer questions, and GitHub Copilot, developed by Microsoft and OpenAI, to help the bank's software engineers be more productive.
The majority of banking organizations have generative AI use cases already in production or have already begun to operationalize them, often focused on customer engagement, risk and compliance, information technology, and other support functions.
“Financial services is a technology-driven business,” says Neel Pardasani, managing director and senior partner at Boston Consulting Group. Though financial services has a head start compared to many other sectors, he adds, the industry can “be a little more cautious about making sure we're doing the right thing and that there are always humans involved in the use cases that are brought to market.”
One early win for the industry has been the adoption of generative AI tools to aid in customer service. “Because of the power of the tools, it’s easy to make the business case work,” Pardasani says.
At Citizens Financial, the new generative AI tool that helps customer reps hasn't reached out to customers directly because the bank continues to test responses and remains wary of hallucinations — responses the AI model generates that are misleading or outright false. “We've gone through those guardrails, but right now we don't feel like we're ready to reach out to customers directly,” Rutledge says.
As big names like Google and Amazon, as well as thousands of startups, launch AI solutions, Citizens Financial said it made a safe bet by leveraging its existing relationships with cloud providers Microsoft and Amazon Web Services, with whom it partners as part of Citizens' hybrid strategy, which also includes Amazon Web Services. “There was a lot of infrastructure already installed, so we were able to build on that,” Ruttledge says.
North American banks are leading the way in AI innovation, with JPMorgan Chase, Capital One and Royal Bank of Canada at the forefront of AI innovation, according to a recent report from Evident Insights. Banks on the continent accounted for more than 80% of all AI research published in the field last year, and Capital One and Bank of America dominate the AI patent space, accounting for two-thirds of all patents filed in the 12 months ending in June 2022.
“There's a lot of scaffolding we need to build on,” says Prem Natarajan, chief scientist and head of enterprise data and AI at Capital One. “We're in a great position to build on that history, but let's also be humble and recognize that this is now a new technology, and given its power, it deserves some respect in terms of taking a test-and-learn approach first.”
In March, Capital One announced a partnership with Columbia University, pledging to invest $3 million in an AI innovation center aimed at accelerating research, but doing so responsibly. “This transformation is going to be so big, and the benefits are going to be so big,” Natarajan said. “And that's OK. They're going to be around for a while. So let's take the time to figure this out and get it right.”
Natarajan said Capital One's approach is to first understand the different use cases the company can explore with generative AI, then determine whether it has control over the data that goes into the model. From there, Capital One can figure out what it can build to test and learn, mitigating unexpected outcomes.
“We want to take an intentional, human-in-the-loop approach from the beginning, so no information is ever leaked out,” Natarajan says.
Visa recently announced three new AI-powered risk and fraud prevention tools for the payments company's corporate clients, and the company prevented $40 billion in fraud last year, nearly double the amount from the previous year.
Visa said the new products are expected to launch in the first half of 2024, and that its tools will include Visa Deep Authorization, a new risk scoring solution aimed at better managing payments made when a physical card is not present.
“We're applying deep learning AI models, and we've trained those models on literally millions, billions of transactions,” said Anthony Cahill, global head of Visa's value-added services. “We provide an informed view of whether it's a proper payment or whether it's actually a payment that requires deeper scrutiny.”
Last year, Visa announced a $100 million venture fund for generative AI startups. Chief product and strategy officer Jack Forestell said that while much of generative AI “has historically been focused on tasks and content creation, this technology will soon not only transform the way we live and work, but also meaningfully change commerce in ways we need to understand.”
Mastercard is also investing heavily in fraud prevention, spending $7 billion on cybersecurity over the past five years, acquiring new technology and working to develop AI tools to make it easier to identify fraud. Mastercard is also investing in about 20 startups to get an early look at new security tools that the company might want to leverage to help prepare for the future of fraud prevention.
“Security permeates our business, from protecting our systems themselves to the new features we offer to our customers,” said Ed McLaughlin, Mastercard's chief technology officer.
