The pace of developments in generative artificial intelligence has put bank and credit union managements into a three-way squeeze.
On one side, they face the pressures of simple competitive reality. “You can’t afford to just marinate on the GenAI topic for a couple of years, to see what happens, because you will be so far behind by the time you begin that you might not be able to catch up,” says Lee Wetherington, senior director of corporate strategy at Jack Henry, in an interview with The Financial Brand.
If competitors act first, “you’ll find yourself in a deep competitive hole very quickly,” adds Wetherington.
On another side, banking leaders face the need to continually maximize their institution’s efficiency, which backers of GenAI promise the technology can deliver. Wetherington points out that if the promise is met, GenAI will reduce the need to hire as many employees to help banks continue to scale.
But the third pressure, shared by banks and credit unions, is the haziest right now: Regulation and supervision.
GenAI Policy: Long on Caution, Short on Public Specifics
As institutions large and small have been venturing into GenAI, regulators have had little to say — publicly, in any case — about GenAI other than issuing general cautions.
For example, Acting Comptroller of the Currency Michael Hsu pointed out in a mid-2023 speech that GenAI models require training.
“While this is part of their magic,” the regulator said, “it also creates a fundamental problem: Since AI systems are built to ‘learn,’ they may or may not do what we want or behave consistent with our values.”
More generally, Todd Harper, chairman of the National Credit Union Administration, said in a fall 2023 speech that AI demanded responsible adoption as it permeates everything from member communication to loan underwriting. “The NCUA is keeping a close eye on developments in this field and measures adopted by our peer regulatory agencies, so the credit union system can continue to innovate while remaining faithful to its statutory mission,” said Harper.
What’s missing, even from a long fall speech by Federal Reserve Gov. Lisa Cook devoted solely to AI, is any clue where the regulators are when it comes to promulgating rules about use of GenAI.
There have been general statements over time about institutions ultimately being responsible for actions resulting from misguided AI — such as discriminatory lending — but the regulators have been issuing such advisories for years, long before ChatGPT and other GenAI hit the headlines. It’s possible that more specific advice or direction is being issued in the field on a case-by-case basis as regional examiners and supervisors work with individual institutions.
Why Most GenAI is in the Background So Far
“Regulators are always one, two or three steps behind whatever’s available,” in the emerging tech area, says Wetherington. “They don’t yet have the tools or the infrastructure to know how to vet the models that are behind any given GenAI application that a bank or credit union is bringing to bear.”
Banks and credit unions are only just beginning to learn about using GenAI, Wetherington continues, so he thinks it will be some time before they begin allowing GenAI and customers to interact without participation or supervision by a human.
“In the interim, there are plenty of things to do with GenAI in the back office and in the assistive context, where you’ve got a human who is curating whatever output GenAI is providing in any context,” says Wetherington. Augmentation, rather than replacement, of humans in banking relationships makes more sense.
Wetherington has been present in private discussions among regulators about GenAI and its potential regulation for banking use. In his judgment, he says, institutions can expect regulators to require proof that GenAI models are not digesting and then disseminating any of customers’ personal financial information. He says it’s possible that large tech providers, such as cloud operators, will need to indemnify financial institutions to give regulators assurance on this matter.
The regulators are on their own learning curve about GenAI, says Wetherington.
In a presentation at a conference late last year by Microsoft and FedScoop the Fed’s own chief innovation officer, Sunayna Tuteja, spoke generally about the central bank’s efforts to set up a GenAI incubator for its own development purposes. Tuteja previously worked at TD Ameritrade, where incubators are used to develop ideas on how to use new technologies, she told listeners.
Wetherington says that he has heard that federal bank regulators are tinkering with GenAI to enable the software to learn how to generate bank examination reports. He says that the idea is for the GenAI to eventually be able to draw on language proven over many, many exams to elicit the actions and responses from regulated institutions that the agencies want to see.
Wetherington says that agencies typically take a long time to respond to market developments. Case in point: Banking-as-a-service arrangements were fairly commonplace by the time federal regulators began commenting about them in speeches.) However, he suspects things may be different this time.
“I think the pressure on regulatory agencies to provide definitive [GenAI] rules is going to be so great that they’re going to have to compress the timeline that we would normally see.”
— Lee Wetherington, Jack Henry
Read more about artificial intelligence in banking:
- In The Wake of Google Gemini’s Chatbot Debacle, An Object Lesson for Banks
- Why it’s Time for Banks to Hire a Chief AI Officer — and What That Looks Like
- Lessons from Ally’s Experiment with GenAI Marketing: Commit, But Carefully
The Potential for AI and Open Banking to Work Together
One of the key areas that GenAI and AI overall can help institutions with is open banking, according to Wetherington. He believes the open banking rules proposed in 2023 by the Consumer Financial Protection Bureau, which are being finalized at this writing, will open up much data that is currently isolated. This is because of American consumers’ widespread fragmentation of their banking relationships across many institutions and financial apps.
With open banking facilitating more connections among consumers’ financial relationships, they will have a better sense of their finances. Wetherington thinks artificial intelligence will help pull all the data together in a coherent fashion.
Wetherington also foresees GenAI leveling the playing field a bit between large financial institutions and smaller ones.
“On the surface, big banks have more data and more budget to buy computing to pore over AI models with the data,” says Wetherington.
However, he says, much of AI now has the ability to benefit from “transfer learning.” This refers to the capability of applying training from past tasks to future tasks. Smaller institutions can benefit from this aspect of the technology, he says.
“Banking is ripe for transfer learning efficiencies,” says Wetherington.