AI in Finance Progresses but Faces Delays Before Widespread Use

Key Takeaways

• AI technology is advancing rapidly in financial services, with productive use cases beginning to emerge.
• Agentic AI (small software programs that learn, reason, and act independently) is the latest phase in AI development for banking.
• Widespread adoption of client-facing AI agents is still 1-3 years away, with smaller institutions lagging behind multinational banks.
• Demo glitches and execution gaps highlight the nascent stage of agentic AI solutions.
• Evident AI Symposium emphasized the need for proper governance, testing, and employee engagement in AI rollouts.

Current State of AI in Banking

AI is already being used for internal processes like fraud detection (e.g., Bank of America’s Erica chatbot, which processed 2 million client interactions in 24 hours). However, public-facing AI agents are not yet consistent or reliable enough for prime time.

Challenges and Delays

• Technical glitches: A recent Demo Day by Liminal Corporation showed multiple failures in AI agent demonstrations.
• Process fixes needed: AI cannot fix broken workflows—banks must optimize processes first.
• Governance and testing: Proper rules and flexibility must be established before deployment.

Expert Insights

• Hari Gopalkrishnan (Bank of America): AI is beyond experimentation but requires clear business problem-solving.
• Manuela Veloso (JP Morgan Chase): Banks should start small with AI and expand gradually.
• Jodie Wallis (Manulife): AI must be a strategic priority for CEOs to drive transformation.

The Road Ahead

While AI adoption is accelerating, consistent high-performance agents for client banking experiences are 3-5 years away. The focus now is on internal efficiencies, with external applications like investment advisory tools coming later.

For more details on bank AI rankings, see Evident AI’s 2025 Index.