The news: Seventy percent of banking leaders believe agentic AI will “have a significant or game-changing impact on the banking industry,” according to an American Banker survey commissioned by SoundHound AI. But only 18% said their bank was “actively deploying” agentic AI, while 33% “are testing through pilot(s).”

Here’s how it works: Agentic AI completes tasks autonomously based on a user’s instructions, rather than just extracting and organizing information as genAI does. Use cases include customer self-service, employee copilots, and fraud and risk management. As we explored in our October 2025 report GenAI and Agentic AI in Banking 2025, banks are currently focused on internal applications of agentic AI. Compliance, costs, and inertia are holding back customer-facing applications.

Zoom out: Uncertainty is the operative issue. In the US, consumer protection, data security, and privacy regulations apply to AI, but no federal regulations explicitly address the development of AI tools. The Trump administration appears to be permissive: An executive order issued in January directed the government to revise or rescind Biden-era rules on AI, and “America’s AI Action Plan” published in July implies tolerance toward AI development.

Our take: For banks to move agentic AI tools from the pilot stage to wide rollouts, they’ll need a business case, a strategic roadmap, and a plan to promptly address tactical issues. Must-haves include strong AI governance, a modern data layer that unifies access to siloed business information, and a framework for customer trust.

Regulatory clarity is also key, though banks that wait too long for it could lose out. Banks’ angle is to balance inherent risks with opportunities. Start with pilot programs for agentic AI, expand them once use cases are proven, and be prepared to move quickly with further development once regulatory guidance solidifies.