Banking on AI: How financial institutions are deploying new tech

Bankers' understanding of AI in its various forms varies, creating numerous opportunities for further education on the distinctions between traditional and generative AI.

While roughly 97% of respondents polled on how familiar they were with AI said they were either very or somewhat familiar with the tech, only 74% had some understanding of the difference between the two main forms of AI. Twenty-six percent had no understanding whatsoever.

This knowledge gap has left C-suite executives scrambling to find qualified talent for the somewhat newly-codified role of chief AI officer and to fill other AI-related roles.

Research departments within large banks such as JPMorgan Chase, Wells Fargo and U.S. Bank are often charged with spearheading research on emerging technology like generative AI, as well as educating others across the organization on potential use cases. 

Those surveyed held that automation and marketing were top-of-mind fields that stand to benefit most from AI.

"We are receiving a lot of interest by our customers over AI, but at the same time, they are cautious about [generative AI] and the risk for hallucinations, which is expected in the financial industry," one respondent said. "I do think that using AI to perform basic tasks is the safest route until [generative AI] can be more reliable."

The tool that is often credited for the recent boom in consumer interest is OpenAI's large language model ChatGPT, which 73% of those surveyed said they have used for either personal or professional reasons. But that figure drops off significantly for similar offerings like Bing AI and Google's Gemini (formerly known as Bard), recording 35% and 20% usage respectively. 

Niche tools including the developer platform GitHub's copilot, Wolfram Alpha and the San Francisco-based AI startup Anthropic's Claude were more obscure at 15%, 11% and 7% of recorded users.

"Probably the most common way that [AI] is used now is ChatGPT to bridge the gap in areas that those in my profession are unsure of, such as coding or coming up with a starting point for ideas," another respondent said.

Asked how they feel about the pace of innovation in AI, more than 60% of respondents felt that both traditional and advanced AI technologies are evolving too quickly, while 37% say change is coming at the right speed. This has made vetting external partners to ensure proper frameworks for governance of increased importance.

"It is important to note that the use of AI in banking is still in its early stages, and banks should carefully evaluate AI vendors based on their specific needs and requirements," John King, a partner in the business transformation practice at Lotis Blue Consulting, said in an American Banker opinion piece. 

That hasn't stopped those who consider themselves "fast adopters" from forging ahead, as 56% are already using AI or expect to use it soon for work or personal reasons. 

Respondents were more willing to trust AI in their personal lives with uses ranging from predicting car or household maintenance needs (57%) to providing financial recommendations for investing or budgeting (56%). Tasks that fell lower on the list include matchmaking with romantic partners (12%) and finding caregivers for loved ones (9%). 

When it comes to work, surveyed bankers mainly trust AI to assist humans — in the form of copilots.

Helping both employees and customers with standard queries were the top two use cases that banking professionals were confident that AI could be mostly-to-wholly responsible for, respectively, garnering 72% and 69% of responses. Fraud detection, report generation and cybersecurity also ranked high on the list.

Banks and credit unions alike have employed AI-powered chatbots for consumers to help relieve the pressure on customer service representatives and free up time for more complicated requests. Use cases for employees specifically, however, haven't seen the same level of adoption.

While 18% of respondents that have access to chatbots use them for transcription services, an equal percentage of bankers that also have access choose not to use the tools. Similar disparities were seen throughout this prompt in areas like navigating benefits, managing reimbursements and employee onboarding.

Respondents were polled on how (if at all) they employ generative AI in their day-to-day lives, as well as the general sentiment and pace of adoption within their organization.

Generative AI is classified in the report as technology that "leverages large language models and describes algorithms (such as ChatGPT) that can be used to create new content, including audio, code, images, text, simulations and videos."

Roughly 61% of bankers said they use generative AI in their personal and/or work life. While that statistic is promising on an individual basis, results for adoption at an institutional scale are more staggered.

More than half of global and national banks surveyed are implementing gen AI at some level over the next 12 to 18 months, recording significant progress in the more than $100 billion asset class. Roughly 40% of regional banks with less than $100 billion are also adopting gen AI.

Citi is one such example. The bank currently plans to have a roadmap to deploy the GitHub Copilot for all 40,000 of its developers by mid-April, as well as other applications for updating legacy software and composing initial drafts of compliance assessments.

"I do believe it's a technology that will, in a sustainable way, have a long-term impact on how we do work for a couple of decades to come," Shadman Zafar, chief information officer of personal banking and wealth management at Citi, said in an earlier interview with American Banker.

Credit unions are also eager to add gen AI tools to their tech stacks, but on a smaller scale when compared to banks. Approximately 13% are focusing on small-scale implementations for specific purposes, and 27% are incrementally wading into the space with individual pilot projects. 

Community banks, those with less than $10 billion of assets, recorded the lowest percentages of gen AI adoption at 28%. The majority of respondents said they are still in the exploratory stage and researching different products, totaling 61%.

Much like traditional AI tools, general use cases for gen AI revolved around improving office productivity and helping fight fraud with 45% and 36%, respectively. Customer-facing applications focus more on marketing communications and helping customer service employees answer questions.