AI roundup: Where banks are implementing AI vs. where they want to be

The trend: The majority of financial institutions (FIs) have integrated AI into their operations, per The Harris Poll. But genAI is still far from being as consistently reliable as banks want it to be, and some hope to get it there sooner.

Banks’ ideal AI capabilities: Major gaps in the technology still prevent banks from entrusting it with more tasks. According to an American Banker analysis, bankers’ AI wish list includes:

• Improved reliability: AI models still sometimes generate nonsensical information. For example, when prompted questions about nutritions, Google’s AI Overview feature suggested eating rocks could be beneficial for users’ health. Banking users need to rely on AI’s ability to say so when it doesn’t have an answer, and to sift out poor-quality sources of data—especially data that includes bias.
• Customizability: FIs are looking for ways to differentiate themselves from competitors, so they want installations trained to make full use of their unique customer data. That way, they can provide a uniquely helpful banking experience that their competitors can’t.
• Data privacy and security: Banks can’t share customer data with just any genAI solution due to compliance and customer privacy requirements. Ideally, AI would forget some interactions or parts of interactions and not retain any information post-query to prevent data persistence and potential misuse.

Current state of AI: FIs remain highly divided on how much time and money they’re willing and able to dedicate to AI implementation. 

For example, JPMorgan and Capital One are leading AI research, development, and staffing. But JPMorgan just gave us a clearer picture of its AI strategy at the Bernstein Strategic Decisions Conference.

• JPMorgan has at least 400 use cases for the technology, but hopes to double that by the end of the year. And it has about 200 dedicated in-house researchers and 2,000 other employees to make that happen, per Efinancialcareers.
• Jamie Dimon told attendees: "We use [AI] for prospect, marketing, offers, travel, note-taking, idea generation, hedging, equity hedging and the equity trading floors, anticipating when people call in what they're calling it for, [and] on the wholesale side, answering customer requests."
• And the bank plans to release MoneyBall, a new AI solution that aims to help portfolio managers better understand the market and their decisions based on about 40 years of data, per The Financial Times. 

Next steps: Not every FI has the budget and the talent to make this happen. Many are still in the early stages of implementing—or considering—AI, per the MIT Technology Review.

• Even some firms with larger budgets have outdated IT and data structures unfit for modern AI applications, which can make the implementation cost-prohibitive.
• And there’s a major shortage of AI talent, though that’s expected to change as demand leads people to the field. 

Key takeaways: Regulators have had a difficult time keeping up with this technology, which makes it risky territory for FIs.

• The European Securities and Markets Authority (ESMA) just said European banks can use genAI, but warned that they’re responsible for any fallout resulting from its use, per Reuters. 
• That’s where AI regulations were in the US before President Biden issued his executive order last fall.

And currently, there isn’t a big selection of genAI companies for FIs to choose among when considering third-party solutions. For example, the immense amount of computing power required for AI can limit the entry of new providers into the market.

• Many exploring AI also have concerns about monopolies and reliance on a few key players—along with the prices and policies they’ll offer.