The headlines are constant and ubiquitous. Depending on the publication or post, one can walk away believing AI will either revolutionize banking or create an existential threat to the industry. Will AI-enabled efficiencies make 30% efficiency ratios possible? Or will agentic AI trigger massive deposit outflows as autonomous agents race to the institution offering the highest rate?

With so much noise surrounding AI — much of it generated by AI itself — it can be difficult for community bankers to know where to begin. This article does not recommend a particular AI model or explain how to implement AI to grow a bank. Instead, it offers a practical framework for evaluating AI in a way that is useful and actionable for your institution.

It begins with four simple questions: Who? What? How? When?

Who?

Who at the bank is responsible for the bank’s approach to AI? Without a person or group responsible for AI use within the bank, it is impossible to develop a coherent and consistent approach. Is it the board? A board committee? The CEO? While there is no single right answer, the board retains ultimate responsibility for the bank’s direction. Formally assigning responsibility to one or more people within the organization is foundational to ensuring that the bank’s approach to AI is guided by what is best for the institution and its stakeholders — not by whatever AI-related noise is loudest on any given day.

What?

What is the bank’s policy on AI use? Even if a bank does not have a formal policy in place, it is almost certain that employees are using AI at home, on their phones and likely at work. While AI offers numerous benefits, it also presents significant risks, including the possibility of entering sensitive or confidential information into a public model. A clear AI use policy helps ensure employees use AI in ways that do not expose the bank to unnecessary risk.

How?

How does the bank plan to use AI? More specifically, what is the bank’s strategic plan for AI? Does the bank intend to be proactive or reactive? Will AI implementation focus on evolutionary improvements to operations or a more revolutionary redesign of the enterprise? There are obvious use cases, such as BSA/AML compliance and fraud detection, but what about opportunities to scale the bank without meaningfully increasing overhead?

When?

When should the bank begin asking these questions and considering what AI adoption might look like? Speaking recently on banks’ use of AI, former Acting Comptroller of the Currency Michael Hsu — not someone known for encouraging banks to take cavalier risks — observed that AI is one of the rare technologies for which the risk of falling behind may be at least as great as the risk of moving too quickly. Hsu is not alone in that view.

Accordingly, the right time to begin thinking about AI in a deliberate and strategic manner is now. A bank does not need to be first in line to subscribe to Anthropic’s latest cutting-edge model, but the greatest AI-related risk many banks face today is failing to answer the first three questions and, as a result, doing nothing at all.

Conclusion

As banks begin answering those questions, they do not have to do so without guidance. In February, the U.S. Department of the Treasury, in collaboration with the Cyber Risk Institute and several financial institutions, released the Financial Services AI Risk Management Framework. The framework is currently the federal government’s primary AI resource developed specifically for financial institutions and is intended to help banks evaluate AI adoption, identify risks and implement appropriate controls. It provides a practical tool for institutions seeking to develop a thoughtful, risk-based approach to AI.

Access the Financial Services AI Risk Management Framework.

The flood of AI-related headlines is unlikely to subside anytime soon. By asking four simple questions — who, what, how and when — community bankers can cut through the noise, approach AI strategically and position their institutions to benefit from the technology, rather than allowing a more nimble competitor to answer those questions first.