Generative AI’s rise adds to US bank uncertainty

Big banks are scrambling to work out what to do with generative artificial intelligence: how to use it to make some of their people smarter or free up others to do only higher-value tasks, and how to ingest and process data more rapidly, speed up decision making and cut cost

Every bank fears their competitors getting good at AI before they do.

Bay Area venture capitalists have a different warning, though: Lenders are missing the threat from everywhere else. “For banks, when I talk about AI, I tell them: What you should be worried about is what if it works?” Angela Strange, a general partner at Andreessen Horowitz, told me recently.

What if AI in the hands of bank customers automates the dull work of shopping for the best rates on simple financial products at irresistible speeds? It probably won’t be long before generative AI bots can look after our money with utmost efficiency. Great for customers, but not so much for banks. But just because it’s possible doesn’t mean it will be done. There are some heavy caveats to this vision. For instance, it could have deeply detrimental effects on financial stability. I’ll return to these. But the direction of travel is real and tells us a lot about the challenges banks face, especially smaller ones.

Matt Harris, partner at Bain Capital Ventures, foresees an existential threat to banks’ core money-spinner: net interest margin – essentially the difference between what a lender pays on deposits and earns on loans. This margin will come under siege from both sides as soon as people and companies haveAI agents that are authorised to move their money for the best deposit or savings rates and refinance their debts, perhaps even multiple times daily.

“This will be the most viral product in fintech history,” Harris told a gobsmacked crowd at the Newcomer Banking Summit in San Francisco this month. “I will have someone out there shopping on my behalf all day every day. And the American banking system as we know it is over when that happens.”

Generative AI is already very good at reading and writing; it just needs to get better at acting independently on what it knows. That capability could be available within one to three years, Harris told me after his talk.

My first objection: Banks will hate this and fight it every step of the way. And if banks’ funding and assets can appear and disappear minute-to-minute, depending on small changes in rates, that will likely be highly destabilising for individual lenders and the entire system. Harris didn’t disagree, but the principle behind this is supported by regulators at the Consumer Financial Protection Bureau.

Late last year, the agency published a long-overdue proposal to implement the open banking requirements of the 2010 Dodd-Frank Act. At its simplest, open banking is the idea that consumers should be able to easily and readily access all their financial data and allow authorised third parties to do so, too. It has been a fact of banking life in the UK and Europe since 2018 and is being adopted in a string of other countries.

European bankers, such as Ana Botin, executive chair of Banco Santander SA, have complained frequently and loudly about the unfairness of having to give up data to tech companies while getting nothing in return.

The revolution promised by open banking hasn’t arrived in Europe, but it has made one big difference: It forced lenders to develop application programming interfaces (APIs), or gateways through which other software connects to a bank’s IT in order to allow the passage of data and instructions. Without these, fintechs like Wise Plc, the London-listed foreign exchange app, wouldn’t function nearly as well and might not exist.

US banks have been busy developing their own APIs regardless of open banking because increasingly they need the ability to connect to the systems of many institutional or corporate clients.

So, if the technology is happening, what will stop me launching SmartMoneyBot 1.0, other than a total lack of computing skills?

First, the financial stability implications are real and serious. The collapse of four US regional banks and Credit Suisse last year showed just how fast a run can happen in a world of social media and effortless online withdrawals. In the US, last summer’s launch of the Fed Now real-time payments service only speeds things up. AI bots constantly moving money for the best rate could kill a bank entirely by accident without anyone having queried its safety. Regulators will have to examine the implications thoroughly.

The other question is whether consumers will easily trust an upstart AI brand to look after their money. The UK’s and Europe’s experiences suggest maybe not. The advent of open banking had industry experts predicting whizzy new platforms where people would consolidate all their banking and investments, leaving old bank brands and branches behind forever. It hasn’t happened, partly due to a lack of demand: People have been loath to abandon the relative solidity of what they know. But it has forced many lenders to develop better apps and smarter services.

Angela Strange of Andreessen Horowitz says that Gen Z and younger generations, brought up on smartphones, will believe differently, especially with the banking crises in 2008 and 2023 in their worldview. “This generation, rather than trusting a bank, they’re more likely to trust the best engineers to look after their money,” she says.

Maybe? I don’t know (this makes me feel old). One thing I do believe, however, is that another wave of competitive chaos is coming to US banking. And it will be most troublesome for the thousands of small, local banks that can’t afford the technology investment required to stay relevant.

Paul J. Davies is a Bloomberg Opinion columnist covering banking and finance. Views are personal and do not represent the stand of this publication.

Credit: Bloomberg