Where banks should focus AI spending to stay ahead, from Wall Street's AI scorekeeper

• Alexandra Mousavizadeh, CEO of Evident AI, said banks should focus on their "competitive edge."
• The measure of success has shifted from use cases to scale, as banks work toward a new "north star."
• Mousavizadeh said banks need to combine bottom-up and top-down efforts to achieve broad adoption.

AI is no longer optional at banks. The road map, and showing how it pays off, is the hard part.

Alexandra Mousavizadeh, the cofounder and co-CEO of Evident, which tracks AI use in the financial industry, said some AI capabilities are "table stakes" for banks at this point — think back-office functions like reviewing legal documents and routine onboarding tasks. Beyond that, though, Mousavizadeh banks need to double down on their "competitive edge."

"If you've got a big chunk of your business in a certain area, the banks will double down on AI there," she said. For firms with a big wealth management business, that might mean helping advisors better analyze client data; for banks more focused on retail, it might mean prioritizing chatbots and customer engagement.

Banks are pouring billions into AI, and the technology is expected to redefine 44% of the work done there by 2030, according to consulting firm ThoughtLinks. JPMorgan, which has snagged the top spot in Evident's ranking of AI maturity, has invested at least $2 billion into the technology and is rolling out tools across its more than 300,000-person workforce.

Some investors and analysts are starting to wonder about the returns, though. On their latest earnings calls, executives at several bulge-bracket banks fielded questions about when AI-driven productivity and revenue gains will start showing up on balance sheets. Banks are under pressure to show that AI is helping them gain an edge.

Mousavizadeh said banks that have taken a "centralized" approach to technology decisions are often able to move faster and embed AI more seamlessly.

Her comments on focusing on building AI into core areas of the business echo those of Dan Priest, the chief AI officer at consultancy PwC, who previously told Business Insider that companies that took a "crowdsourcing" approach to AI adoption had a "fairly disappointing" return on investment.

Priest said a shift to a "top-down" approach has been more effective, allowing clients to focus on fewer tools and achieve deeper mastery of a smaller set of tasks.

AI agents are top of mind as banks race to prove that their AI spend is worthwhile, but Mousavizadeh said the uptake is still in its early stages, particularly in externally facing roles, like bankers and traders. In those cases, agents will likely be combined with humans for the foreseeable future, as banks, some of the most highly regulated companies, work through what guardrails to install.

For the past six months, Goldman Sachs has been working with Anthropic on co-autonomous AI agents that can automate tasks in the firm, including accounting for trades and transactions, and client onboarding. The bank's tech chief anticipates that the agents will launch "soon," and that it's too early to think they'll lead to job losses, CNBC reported.

How banks are measuring AI wins

Over the past year, banks have shifted how they measure AI success, Mousavizadeh said, moving away from tracking specific use cases toward scaling capabilities. They're thinking about how to apply capabilities in one line of business to others, and build an internal "architecture" that allows AI to reconfigure workflows companywide.

Achieving that scale requires a combination of top-down and bottom-up approaches. Banks need to get technology into every employee's hands, Mousavizadeh said, and mandatory training often yields better results.

But mandates alone likely aren't enough.

"AI is a funny thing. You do need a culture where there's a lot of creativity," Mousavizadeh added.

Looking ahead to the AI-integrated future, Mousavizadeh said that banks have a new "North Star": what does a fully AI-integrated bank look like a few years from now?

"You need to be able to work back from that," she said.

More than applying AI to preexisting products, that means creating new systems in the image of the bank of the future — and doing it faster than your competitors.