“We really work to make sure we have the appropriate guardrails and security in place for our general use cases,” says Michael Ruttledge, chief information officer at Citizens Financial. 

The Rhode Island–based bank says it has taken a thoughtful approach to generative AI, including the creation of a steering committee to ensure employees aren’t going rogue and developing their own projects. The two generative AI use cases that Citizens Financial says will go live this year include an enterprise search tool that helps customer service representatives answer questions and a GitHub Copilot, developed by Microsoft and OpenAI, to help improve the productivity of the company’s software engineers. 

A vast majority of bank organizations are either in production or have gone live with generative AI use cases, often focused on client engagement, risk and compliance, information technology, and other support functions.

“Financial services is a technology-led business,” says Neil Pardasani, a managing director and senior partner at Boston Consulting Group. And while financial services starts ahead of many other sectors, he says the industry can also be “a bit more cautious in terms of really making sure you get it right and the use cases that go into the market always have got a human in the loop.” 

Many early wins for the industry have been in adopting generative AI tools to assist customer service. “It is an easy place to make the business case work, because of the power of the tools,” says Pardasani.

At Citizens Financial, the new generative AI tool that’s helping customer representatives doesn’t directly reach clients, as the bank wants to continue to test the responses that are given and is still keeping an eye out for hallucinations, which are responses that an AI model can produce that are misleading or outright false. “We are going through those guardrails, but for now, we don’t feel that we’re ready to go out directly to a customer,” says Ruttledge.

With major players like Google and Amazon launching AI solutions, as well as thousands of startups, Citizens Financial says it took a safe bet by leveraging the company’s existing relationships with Microsoft and Amazon Web Services, the cloud providers the company works with as part of Citizens’ hybrid strategy that also includes Amazon Web Services. “We were able to build upon that as we already had a lot of the infrastructure installed,” says Ruttledge.

North America banks are leading in AI innovation, according to a recent report published by Evident Insights, which says JPMorgan Chase, Capital One, and Royal Bank of Canada are at the forefront of AI innovation. Banks on the continent were responsible for more than 80% of all AI research published by the sector last year, while Capital One and Bank of America dominated the AI patent landscape and were responsible for two-thirds of all patents filed in the 12-month period ending in June 2022.

“There’s a lot of scaffolding for us to build on,” says Prem Natarajan, chief scientist and head of enterprise data and AI at Capital One. “We are in a great position to build on that history. But let us be humble and recognize that this is now a new technology which, appropriately given its power, deserves its own respect in terms of first taking a test-and-learn approach.”

In March, Capital One announced a partnership with Columbia University and committed $3 million to invest in an AI innovation center that aimed to accelerate research, but do so responsibly. “This transformation is going to be so big and the benefits are going to be so large,” says Natarajan. “That’s okay. They’re going to be around for a while. So let’s take our time figuring this out and do it right.”

Natarajan says Capital One’s approach is to first understand the various use cases the company can explore with generative AI, then determine what data it can control that goes into the models. From there, Capital One figures out what it can construct to test and learn, as well as mitigate any unforeseen outcomes. 

“We want very deliberately to do a human-in-the-loop kind of approach to start with, so that you’re never actually quite exposing things outside,” says Natarajan.

Visa recently unveiled three new AI-powered risk and fraud prevention tools meant for the payments company’s business customers. The company blocked $40 billion in fraud activity last year, nearly double from the year prior.

The new products are launching in the first half of 2024, Visa says, with tools including Visa Deep Authorization, which is a new risk-scoring solution aimed to better manage payments that are made when a physical card isn’t present. 

“We’re applying deep AI learning models, and we’ve literally trained that model with millions and billions of transactions,” says Anthony Cahill, global head of value-added services at Visa. “We’ll give an informed view of a good payment or, actually, is it a payment that needs closer inspection?”

Last year, Visa announced a $100 million venture fund for generative AI startups, with chief product and strategy officer Jack Forestell saying that while much of generative AI “so far has been focused on tasks and content creation, this technology will soon not only reshape how we live and work, but it will also meaningfully change commerce in ways we need to understand.”

Mastercard also has invested deeply in fraud prevention, spending $7 billion on cybersecurity over the past five years, including the acquisition of new technologies and developing AI tools that make it easier to identify fraud. Mastercard has also invested in around 20 different startups to get a first look at emerging security tools that the company may want to use to support future readiness in combating fraud. 

“Security pervades our business, from protecting the systems themselves to new capabilities that we’re bringing out for our customers,” says Ed McLaughlin, chief technology officer at Mastercard.