US banks turn to AI for expense benefits, less for revenue gains – survey

US banks expect the expense benefits from AI adoption to trump perceived revenue growth opportunities, according to a D.A. Davidson survey.

The survey found that US banks project the expense benefits of AI to be about twice the size of the revenue opportunities over the next two to three years. Banks anticipate an average 4.5% to 5% decline in expenses in relation to AI adoption, compared with a 2.5% to 3% increase in revenue.

Larger banks anticipate greater benefits in both categories. Banks with over $50 billion in assets project a 5.6% expense cut, while revenue growth is expected at 3.1%.

Most of the 73 respondents indicated that efficiency was the primary driver behind AI investment. This was not surprising for the D.A. Davidson analysts, as AI use cases tended toward expenses, including Bank Secrecy Act compliance, call center operations and other tasks that can be automated.

The benefits of AI may ultimately revolve around resource redeployment, with banks adopting AI to limit staff increases and reassign existing resources or personnel, the analysts said.

"Our conversations with bank management teams suggests many view AI efficiency benefits as likely to be offset by increased spend on revenue producing headcount," the analyst wrote in a July 8 report. "In this way, expense savings may be characterized as changes to income statement geography, but ultimately drive positive operating leverage as back office/cost center headcount is replaced by revenue producing personnel."

At the same time, cost savings are difficult to calculate, the survey showed. Just 11% of respondents suggested cost saves are measurable, versus 89% saying otherwise.

The survey also found that AI may grant a slight competitive advantage to larger institutions. Additionally, it showed that the main limiting factors for broader AI deployment were data quality, model transparency and regulatory uncertainty.

The majority of the surveyed banks are active in AI adoption, with about 42% providing individual user tools and an additional 35% having identified and implemented quantifiable use cases.

"Our sense is that tangible use cases are accelerating throughout the industry, and while one could say AI adaptation is 'on the come,' we'd suggest it's more accurate to say it's here,'" the analysts wrote.