Banks’ measured approach to AI is warranted. It’s not popular, or even a good idea, to jump in head-first with AI — which itself is a fuzzy term. Banks appear more willing to experiment with AI when it has accepted applications, according to the survey. Dedicated investments, targeted research, and changes to the organization require that banks have a practical understanding of AI, an idea of specific applications, and appropriate guardrails.

According to the American Banker study, bankers may be more comfortable with applications for AI that are familiar, have relatively low risk, or for which the cost of a mistake is relatively low. “Familiar” and “relatively low risk” are related and include the longstanding use of machine learning to do things like detect fraud or money laundering, which over half of bankers are comfortable with; to provide financial recommendations; for underwriting algorithms; or for routing customer concerns from virtual assistants.

Generative AI is a somewhat different story. A shift from machine learning to generative AI is that, instead of detecting patterns, generative AI poses risk by creating content that may be inaccurate, imprecise, or biased. Bankers may be relatively comfortable with generative AI for low-risk applications like generating internal reports and for research, but less so with outward-facing applications. However, given the pace of technological change, they should at least be thinking about future use cases and applications.

Strategic considerations that stem directly from education about AI are, first, the risk and compliance measures that need to be in place and, second, what makes the most sense to try on a small scale. As we wrote in February, financial institutions (FIs) need to be well-informed about AI’s benefits and risks without depending on vendors and set their own well-informed guardrails. Implementation should only come after steps are taken to achieve that state — like a working group for AI, data management and integration, and upskilling employees, to cite examples from the American Banker report.

Based on our analysis, banks’ senior leadership should be laying a foundation for AI that ensures the organization is ready to capitalize on its promise. From there, they should be looking to frame a specific AI strategy, in three parts: