Most observers agree that generative AI is poised to reshape financial services operations from the inside out, but it's trickier to predict how long it will take for banks to see a return on investment in AI — and how much of the investment going on today is being applied correctly.
Venture capital funds pouring into the AI sector drove Goldman Sachs to predict last month that U.S. private AI investment is on pace to double over the next two years, reaching $100 billion by 2025, with Google just days ago announcing plans to invest $2 billion in Anthropic, a rival to OpenAI.
Now bankers trying to plan their own AI investments face fresh uncertainty from President Biden's
This extra scrutiny may cause investors to be more selective in their investments, but Wall Street seems to view Biden's move as having a neutral effect in the long term.
"We wouldn't take this step up in AI scrutiny as a negative for the sector," said Solita Marcelli, chief investment officer for the Americas at UBS Global Wealth Management, in a Tuesday note to investors. With tech giants recently committing to future robust AI infrastructure spending, UBS expects strong AI-sector growth next year, she added.
As things currently stand, bankers are already dealing with the challenge of sorting through overhyped AI products, said Alex Johnson, a fintech researcher and founder of the newsletter Fintech Takes, during a keynote speech at the recent Bank Fintech Fusion conference in Scottsdale, Arizona.
"There is absolutely a bubble in AI right now, with a lot of vaporware, a lot of stuff that's not real, a lot of stuff promising magical outcomes that aren't going to be possible, and I think that a lot of venture capitalists are going to lose a ton of money on AI," Johnson said.
Venture capital firms are so eager to plant a flag in emerging AI projects that many are writing multimillion-dollar checks to entrepreneurs brandishing "three-page pitches," Johnson said, noting that a lot of these investments will be wasted money.
"There are probably 200 people in the U.S. who know how to build these large language models and they work for Google, Microsoft, OpenAI, Apple and Amazon … so if you're not one of those companies, you won't have access to the technical talent needed to build the first level of the stack for generative AI," Johnson said.
Until investors get choosier, banks can focus on the ancillary products — software, application programming interfaces and new technology — that will introduce new workflows and use cases transforming the way banks and credit unions do business, Johnson said. These ancillary products also will likely accrue the most value for investors, he believes.
"It's probably going to take five years — and maybe more — to sort out the investment hype from reality, because apart from tech giants like Google and OpenAI, smaller players are just placing crazy bets right now," Johnson said.
The AI hype cycle is not happening in a vacuum. The attention to AI is happening alongside the growth of open banking, which allows third parties to tie into financial services to enhance their own products.
"My hypothesis is that generative AI and open banking, which U.S. banks are starting to adopt, are the next confluence of trends that will drive the financial services industry forward," Johnson said.
Generative AI is also fraught with limitations, including errors, potential misdirection and bias based on flawed data fed into models, which means banks should plan to move slowly in planning and adopting any AI-powered changes, he advised.
"The combination of open banking and generative AI will define the next age of innovation in financial services, enabling banks to create new products and experiences for customers by finally being able to use bank's existing troves of unstructured data plus all the data they can convince their customers to bring," Johnson said.
Banks can begin planning for these new opportunities now, but they shouldn't worry about whether to invest in AI in the immediate future, he advised.
"No one knows which layer of that stack we're going to see the most value in, but I think a lot of the value will surround APIs, and anyone who tried to go too deep into the AI stack will find themselves competing with these giant tech companies that have vastly more resources to build large language models," Johnson said.
