As some financial institutions race to roll out artificial intelligence, experts warn that they need to figure out how to generate tangible value from the technology over the long run. Cost savings may not be enough, practitioners said at a Finovate Fall conference panel in New York this week.
"Cost savings are a really easy target," said Milton Santiago, global head of digital solutions at Silicon Valley Bank. "When we start looking at driving value, it's harder … driving value is going to be very different for every customer and for every bank."
AI, particularly generative AI adoption, is becoming widespread across financial institutions. In an EY
Defining value
There is a risk proceeding with generative AI without clear business goals, panelists said.
"Don't do it for the sake of using generative AI. Do it because you're going to create value," said Dan Latimore, chief research officer at The Financial Revolutionist. "The classic levers on this are increased revenue, decreased revenue," and risk mitigation, he added.
Cost cutting and efficiency can be considered low-hanging fruit in AI rollouts, but KPIs will depend on the needs of each institution's business goals and client base, noted Santiago.
Sarah Hinkfuss, a partner at Bain Capital Ventures, said an AI tool set should be designed to carry out designated tasks with higher quality results than legacy methods, and for a lower price. As an example, she cited one product that reviews communications for regulatory purposes, reducing the need for humans to take on the task from scratch.
Specialist versus platform approaches
Platform approaches to AI adoption — where one tool can resolve a series of use cases — have their place, but there can be compelling business cases to deploy so-called point solutions — in which one tool addresses a particular need, said Hinkfuss.
"If you try to go too broad, too quickly, you end up being that 'three things to no one,'" she said. "You can't be higher quality, lower cost and faster. You have to be very narrow, especially when you're trying to automate work, and all the agentic applications [autonomous AI agents that perform tasks with little or no human intervention] are doing that."
For Santiago of Silicon Valley Bank, the ways AI tools are deployed are evaluated on a case-by-case basis.
"There may be solutions where I may choose a suite versus a point solution, depending on the value I'm looking to get … every business case has a cost benefit," he said.
The AI race
Multiple approaches to AI deployment — whether in-house or tapping third-party partners — bring a host of opportunities and risks. Larger banks may prefer to develop capabilities internally, but that brings a speed to market risk, suggested Hinkfuss.
"A lot of the biggest banks are actually focused on building it internally," she said. "We still don't know whether that extra level of specialization ... actually gives an advantage [over] an open-source model, for example, and fine-tuning it within your environment."
There can be an argument made that smaller regional banks that top off-the-shelf solutions "are out of the gate faster using proven tools," while large banks focus on differentiation, she noted. Smaller banks' entry into generative AI through third-party platform providers can be done at a significantly lower price point because the third party is doing the research and providing the security and readiness for the rollout, Santiago said.
Smaller institutions, however, could find themselves challenged by cultural issues as they decide how fast they will go.
"You're going to have individuals that are AI curious, you're going to have people who are AI fearful, and then you're going to have those who are just enamored by the technology," said Santiago.
Internal implications
The mainstreaming of generative AI across financial institutions — whether for consumer or internal-facing use cases — will likely enhance expectations among new recruits, especially as a growing number of rote tasks are delegated to AI tools, panelists said. The bar for recruitment of entry-level employees will be higher, but may empower talented individuals to rise.
"The [labor] pyramid will just be much narrower," said Ihar Mahaniok, managing partner at Geek Ventures. "Top talented people with low experience — armed with today's AI tools — could be extremely valuable to a company … they will be the right people to actually grow into the senior [ranks]."
