What will artificial intelligence bring to the financial services industry? This question is becoming less crucial than how financial institutions are approaching AI implementation, development and innovation. 

Seventy-three percent of C-level bank executives are interested in or already using AI tools within their institutions, according to a CCG Catalyst Consulting report that surveyed 108 C-level bank executives between May and June of 2023. 

“What I am seeing is bankers having a lot of enthusiasm, a lot of interest around the efficiencies that [AI] can drive — but they’re struggling to understand how to capitalize on those opportunities because there’s just so much noise out there right now,” Kate Drew, director of research at CCG Catalyst Consulting, told Bank Automation News. 

FIs are taking varying approaches to implementing AI. But they share the goal of promoting better, faster, cheaper and more scalable operations, Katya Chupryna, head of fintech strategic investment at Citibank, said at the Finovate Fall 2023 event in New York. 

Depending on resources and capital, banks are considering building, buying or both. 

“Many large organizations are probably looking at both, buy and build, and it’s a matter of understanding what makes sense to build internally versus what makes sense to buy,” Jennifer Warren, head of markets digital strategy at Barclays, said recently during BAN’s “Global ideas for better banking AI” webinar. 

When to build

Some financial institutions argue that desired control and sensitivity are reasons enough to build in-house. 

“You want to build if it’s something around sensitive data, if it’s highly customized or if it’s part of your core business,” Robert Bole, managing director of the digital practice group at $1.5 trillion Sumitomo Mitsui Banking Corp., said at Finovate Fall. 

TD Bank’s Jo Jagadish, executive vice president and head of corporate products, agreed: If a product is “really core to your business model” it makes more sense to build it. However, if you want to scale quickly — buying is your best bet, she said. 

Barclays’ Warren said: “It really comes down to your resources and skill sets and how much control you want to retain, how much intellectual property you want to build into it.” 

When to buy

Toronto-based Scotiabank is letting the technology experts take the lead on answering the buy-or-build question. 

“It is absolutely ridiculous to think of a technology company setting up their own bank for their own financial service. … Why would a financial services company pretend to be a technology company and build their own technology?” Grace Lee, chief data and analytics officer at Scotiabank, said at the Sibos event in Toronto last week. 

In fact, FIs are purchasing out-of-the-box generative AI solutions from well-funded startups or large technology companies that offer pre-trained large language models, Prag Sharma, global head of AI at Citi, said during the recent BAN webinar. 

While the purchase is being made to kick start generative AI, there is a need for collaboration to ensure the technology can work with the bank’s data and perform for the organization — that’s where a combined effort of buying and manipulating the technology comes in. 

When to co-engineer

Financial institutions including Barclays, Citi, Deutsche Bank, HSBC and Scotiabank all are using AI — and all have third-party vendors by their sides. 

“Build versus buy, from my perspective, is no longer build versus buy, it’s all co-engineering,” Inwha Huh, managing director at $1.4 trillion Deutsche Bank, said at Sibos 2023.  

Building technology in house, whether for cloud infrastructure or new solutions, is both time-consuming and expensive. FIs and fintechs recognize they can pool their resources when it comes to AI to save money and get to market more quickly. 

“We are looking at co-engineering as a hybrid,” Huh said, noting that purchasing a fintech solution doesn’t cut it anymore; rather, the bank needs its own base of AI knowledge. 

Whereas traditionally, fintechs could install the technology and manage training and updates, AI now is gaining traction throughout the finance industry and the banks must have the skill sets in place to approach it, Huh said.  

At Deutsche Bank “we are heavily working on that and developing the skill set to understand how to develop large language models,” she said. 

Although bank teams are aiming to put AI training into place, there is still a need for external solutions, she said, adding “we understand that the build will take too long.” 

Similarly, JPMorgan spends billions of dollars innovating solutions in-house but is also increasingly looking to “partnering with fintechs to develop new solutions,” Raymond Nazloomian, executive director of innovation, partnerships, corporate development at JPMorgan Chase, said at Finovate Fall 2023. 

The $3.7 trillion bank has a dedicated team for fintech partnerships and understands that “it’s really going to be difficult for any company of any scale to build everything themselves and always be at the forefront of innovation,” Nazloomian said. 

Financial institutions need to figure out “what are domains that we think we want to partner with … and then what are domains that we might apply or invest in,” Nazloomian said. 

Selecting a third-party vendor

Once a bank determines that it’s looking to a third-party tech provider for AI implementation, it must select a company that understands the pain points to be addressed, EJ Achtner, global head of AI for commercial banking at HSBC, said during a recent BAN webinar. 

“It really boils down to, what do you need to do to solve for your customers,” he said. 

Barclays’ Warren agreed, saying, “When you look at partners, you want someone who really wants your business, who wants to collaborate, is going to provide that kind of customer success and support and really help you achieve your use cases.” 

Along with complementary business strategies, a fintech must also be able to scale with the bank, Achtner said. If a fintech doesn’t understand the language of models and banking — it won’t be able to scale. 

Citibank, too, takes a mixed approach to buying, building and partnering with fintechs for its AI technology. 

“There are certain places where you want to partner with very early stage companies who are really pushing the envelope, and there are other parts of the ecosystem where you actually want to partner with more mature companies,” Citi’s Chupryna said. 

When choosing a fintech partner, due diligence is vital for FIs to evaluate whether the founder and company are mature enough for “prime time,” she added. 

“Sometimes, if a company is just not ready for prime time, you don’t partner with them,” Chupryna said. “But you always kind of keep watching and figure that perfect stage when they’re ready to come in and really help you change your business in a positive way.” 

AI in action

Whether financial institutions are buying, building or collaborating on their AI-driven capabilities, several solutions are already in the market today. 

Bank of America’s AI-driven virtual assistant, Erica, was developed in-house by the $3.2 trillion bank in 2018. More recently, the bank’s proprietary Erica technology was integrated into its commercial bot, bringing an AI-driven customer experience to corporate and commercial clients.  

Using in-house technology through Erica allows Bank of America to consume customer feedback and act on developing the functionality further based on client interaction, Chief Executive Brian Moynihan said at Sibos.  

Meanwhile, $1.9 trillion Wells Fargo launched its AI-driven chatbot, Fargo, in collaboration with Google Cloud, in October 2022.  

On the buy side, NatWest recently partnered with Amazon Web Services to use the tech giant’s generative AI capabilities to help nearly 10 million people manage their finances. The $886 billion bank aims to provide personalized services, including saving, buying a home, education on money management and establishing or growing a business by using generative AI. 

Similarly, HSBC is using AI to enhance developer productivity and speed products to market more quickly. The $2.9 trillion bank is also using Google Cloud’s AI model to fight money laundering with a feature that is able to reduce alert volumes by 60% while lowering the number of false-positive incidents.  

Contributing to the AI ecosystem

Financial institutions are coming to their own conclusions on whether to buy, build or combine strategies, and as they decide, one thing is certain: The AI space continues to be uncharted territory. 

“There are no experts in this space,” HSBC’s Achtner said. “We’re all learning, we’re all testing, and I have yet to find a magic organization out there that can solve everyone’s problems.” 

To advance AI within the financial industry, banks can learn from one another and share their findings, Achtner said. Open-source tech is one way to do so. 

Citi’s Sharma echoed that opinion: “There’s huge benefits to utilizing open-source technologies in the right manner.” He added that the industry can learn from Meta, university platforms and other entities that open-source their technology. 

In addition to advancing AI, open-source also provides a cost-effective avenue for AI implementation, Barclays’ Warren said. Bringing in open-source models offers opportunities for FI to create their own IP on top of the technology and “I think that’s the future,” she said. 

While banks are considering whether to buy versus build, the question now may be: How can a financial institution not only consume but contribute to the evolution of AI within the industry, Sharma said. “We have to be thinking about how we can do that.”