Banks are working to improve their customer experience using artificial intelligence, but challenges remain.
At American Banker's Digital Banking conference this week, executives from BMO, Truist and SAP said financial institutions are moving quickly to deploy AI tools aimed at making banking simpler, more personalized and more responsive.
While a recent American Banker survey found that 57% of U.S. bankers expect AI to have a considerable impact on customer experience within two years, this isn't fast enough, some speakers said.
"Two years is too long. Change is happening right now," said Harveer Singh, chief data officer for consumer and small-business banking at Truist.
Singh said customers expect banks to anticipate their needs based on predictable patterns, like paydays and bill cycles, and deliver useful advice in the nick of time.
Truist's
Brianna Elsass, head of digital experience and technology for BMO's retail and small-business segments, said AI needs to blend seamlessly into customer interactions.
"Customers need to feel like they're interacting with humans, not robots," Elsass said. "We're focusing on using AI to enhance omnichannel experiences — ensuring customers have consistent, personalized service whether online, in a branch or on the phone."
AI is helping BMO transition from traditional chatbots to more sophisticated systems that understand customer history and preferences, enabling frontline staff to deliver better service, Elsass said.
Chatbots are a common form of AI in banks; according to American Banker's survey, nearly 80% of organizations have chatbots either fully implemented or in pilot or rollout.
But Erin Holloway, executive advisor at SAP, said chatbots are the least favorite way to interact with a bank.
"They're absolutely getting better, but half the time, you can't get the information that you need. And it's not the chatbot's fault. It doesn't have access to the data," Holloway said. "AI is only as good as the data it has access to, that it's built on, and it has to be real, it has to be reliable, and it has to be relevant and fresh. Once it goes stale, it's no good anymore."
Banks are just starting to offer agentic AI to their corporate customers, Holloway said. JPMorganChase recently signed an agreement with SAP through which it's embedding the vendor's agentic AI technology into its asset management platforms.
An AI agent might say, "I see that the balances in your Norwegian account have gone low. The most efficient way for you to bring that up to where it needs to be is to make a debit of $10 million from your multicurrency account in London. Would you like me to go ahead and start that transaction?" Holloway said. "So it's doing half that work for you, you just have to say yes or no." Rules and regulations can be embedded in what the AI agents do and tracking mechanisms can keep those guidelines in place.
Both Singh and Elsass stressed the need for AI systems to work invisibly, delivering personalized experiences without making customers feel like they're interacting with a robot.
"It should not look like AI," Singh said.
Elsass said that while back-end AI can drive efficiency, customer-facing applications need to feel human to help build trust.
Challenges to deploying AI using customer data
One of the biggest hurdles to AI-based personalization is legacy infrastructure, surveyed bankers and conference attendees agree.
"Our systems weren't built for the kind of cross-platform data integration AI requires," Elsass said.
Regulatory compliance is another major concern as banks experiment with AI. "You need to pick your use cases carefully — not every shiny object should be chased," Singh said. "Test, learn and pivot quickly."
Choosing AI providers can also be difficult, Singh said.
"When I arrived in the U.S. 25 years ago, for the first time I went to a grocery store and I looked at the aisle that had ice cream, I was like, How do I choose? I think that's exactly what AI is right now," Singh said. "How do you choose all these tools that are out there, all these models, all these vendors out there trying to sell you something? You will end up trying a few and discarding some."
The bank's existing cloud platform and consulting partners will help dictate some of these choices, he said.
Singh said Truist will continue its work of centralizing customer data to fuel AI-driven personalization, while Elsass stressed the importance of preparing staff at all levels to understand and apply AI technologies.
Holloway said, "We've just barely scratched the surface on what can be."
"I think that we are getting there," she said. "Regulators have to be more supportive of embracing it with us and meeting our partners in this and knowing that it's the best thing for our customers."
