In the global AI ‘arms race,’ banks are stretching their tech experimentation

The news: Despite their wariness around using third-party artificial intelligence (AI) tech, banks are investing heavily in the space. Retail banks are forecast to spend $4.9 billion on AI platforms by 2024, growing 21.8% compounded annually from 2019, per GlobalData research.

The AI arms race: The release of Open AI’s ChatGPT in November 2022 started what’s being described as an AI arms race, per Bloomberg.

  • Inquiries about AI from banks jumped five-fold in Q1 2023 compared to the same period a year ago, according to Eigen Technologies, which helps firms including Goldman Sachs and ING with AI.
  • Data from the consulting firm Evident, cited by Bloomberg, shows that banks leading in their adoption of AI are allotting 40% of all open job roles for AI-related hires, such as data engineers and quants, as well as ethics and governance roles.
  • JPMorgan advertised globally for 3,651 AI-related roles from February through April, Evident’s data showed—almost double its nearest competitors, Citigroup and Deutsche Bank.

Banks are betting the tech will help personalize their interactions with customers and simplify and speed up complex risk analysis and modeling. Here’s a look at how some large financial institutions (FIs) are exploring the efficiencies that AI promises, per Bloomberg

Within the US:

  • Wells Fargo is using large language models—AI networks trained by text scraped from the internet—to help determine what information clients must report to regulators and how the bank can improve its business processes. It’s also used Google Cloud’s conversational AI platform, Dialogflow, to build a chatbot for customer assistance.
  • Morgan Stanley says it’s allowing businesses around the firm to run tests using open-source large language models. In April, the bank said it had patented a model that seeks to detect the direction of monetary policy by using AI and deep learning to classify communications from the Federal Reserve as “hawkish” or “dovish.”
  • JPMorgan filed a patent application in May for a ChatGPT-like service to help investors select particular equities, per Bloomberg.

Outside the US:

  • BNP Paribas is using chatbots to answer client questions while AI seeks to detect and prevent fraud and money laundering.
  • Societe Generales Cast scans for possible misconduct in capital markets. It operates in 26 languages to process 2.5 million hours of conversation and 347 million emails each year, the bank said.
  • Klarna Bank CEO Sebastian Siemiatkowski told Bloomberg TV that all employees at the Swedish buy now, pay later fintech are offered a ChatGPT-4 account and encouraged to experiment with the new tech.
  • ​​Deutsche Bank is deploying deep learning to analyze whether international private banking clients are too heavily invested in a particular asset, and to personalize matches with suitable funds, bonds or shares. Human advisers vet and forward AI-generated recommendations.
  • Barclays CEO CS Venkatakrishnan said it’s still in the “serious study” phase. It’s looking into using AI to give customer service agents a better overview of client finances, but doesn’t expect to implement AI tools across the firm for several years.

The risk/reward calculation: Bankers can’t afford to disregard their fiduciary duty not to trade on unreliable information. And unfortunately, that’s still a good description of what AI provides. OpenAI’s interface for ChatGPT includes a disclaimer: “ChatGPT may produce inaccurate information about people, places, or facts.”

In a report, OpenAI’s researchers further elaborated on this warning. “Even state-of-the-art models are prone to producing falsehoods—they exhibit a tendency to invent facts in moments of uncertainty,” they wrote. “These hallucinations are particularly problematic in domains that require multistep reasoning, since a single logical error is enough to derail a much larger solution.”

Banks that are identifying areas where AI can create efficiencies and investing in hiring and training staff will also need to redesign risk frameworks to deal with this danger. They’ll also need to look into intellectual property considerations and continue tracking the still-evolving AI regulatory environment.

Dig into what the technology could mean for banking in our report, ChatGPT and Generative AI in Banking.

This article originally appeared in Insider Intelligence’s Banking Innovation Briefing—a daily recap of top stories reshaping the banking industry. Subscribe to have more hard-hitting takeaways delivered to your inbox daily.

"Behind the Numbers" Podcast