Banks are taking a holistic approach to generative AI integration this year with nearly 80% of FIs projected to adopt enterprisewide gen AI strategies by the close of 2025.

That number was provided to Bank Automation News by Connor Loessl, associate partner for global banking and financial markets at IBM Consulting.

However, only 8% of banks were using gen AI across all enterprise functions in 2024, while 78% said they didn’t have a companywide gen AI strategy, according to IBM’s 2025 Outlook for Banking and Financial Markets, published Feb. 5.

Key applications

Four use cases are continuously pushing gen AI adoption, Loessl said:

• Client experience;
• Operations transformation;
• Risk management; and
• Modernization.

Areas such as automated software for testing bank applications and converting checks into electronic payments also present high potential for gen AI integration, he said.

IBM’s 2025 outlook report combines data from a 2024 CEO survey of 3,000 chief executives across industries and a 2025 study of 600 banking executives at small- and medium-sized enterprises. Data from the two studies was supplemented by IBM quantitative analysis of annual reporting for about 600 global banks with assets above $10 billion, Loessl said.

Scaling integration

Three areas should be considered when implementing gen AI in banking, he said:

• Smarter learning models;
• Governance; and
• Scalability.

Financial institutions are finding cost-effective ways to scale gen AI integration with the emergence of smaller language models and AI-powered tools that enhance large language models, such as the open-source AI InstructLab platform from IBM and its subsidiary Red Hat, Loessl said.

For example, IBM is working with a large bank client this year to drive the bank’s goal of generating at least $1 billion in efficiency gains through gen AI integration, Loessl said, without disclosing the bank’s name.

When IBM initially met with the bank, the FI was operating more than 100 siloed gen AI-powered pilot technologies that were all promising but lacked centralized governance for how they’d be used throughout the bank, he said.

A cohesive approach to gen AI integration — rather than adding pilot technologies for isolated functions — is critical for banks to optimize return on investment, Loessl said.

“If we think about things like the cost of inferencing and banks’ ability to capture the value of generative AI, a siloed pilot is not going to get them there in terms of their goals for productivity and cost takeout,” he said. “There’s really this importance around developing as-a-service generative AI capabilities centrally and dispersing them across their business units.”

Banks are also looking to agentic AI for applications such as marketing, customer engagement and content creation, Loessl said, adding that he expects to see an uptick in agentic AI alongside gen AI this year.

Bank implementation

IBM works with $952.7 billion, London-based NatWest on gen AI applications, Loessl said. Through the collaboration, NatWest has developed a digital lending assistant as well as Cora+, a digital assistant that answers bank inquiries.

NatWest is also investing in Serene, an early-stage AI platform that detects early signs of financial distress, the bank announced Jan. 28.

Other big banks IBM works with include:

• $3.9 trillion JPMorgan Chase;
• $2.6 trillion Bank of America;
• $2.4 trillion Citi; and
• $531 billion Truist.

Register here for Bank Automation Summit 2025, taking place March 3-4 in Nashville, Tenn. View the full event agenda here.