Banks are looking to 2026 as a transformative year toward becoming fully AI-powered. Currently, the integration of AI in banking is hindered by fragmented data systems, compliance pressures, outdated legacy technologies, and institutional resistance to change. Many AI initiatives remain confined to isolated proofs of concept, marked by poor governance and inconsistent results.
Read More
Bank executives face mounting expectations for productivity and are under pressure to deliver tangible outcomes. Despite significant increases in AI budgets over the last two years, most U.S. banks have achieved only sporadic tactical successes rather than comprehensive strategic transformations. A review of the top 40 U.S. banks indicates that current AI efforts are predominantly reactive and siloed, yielding varying degrees of value.
To advance AI integration, banks should establish a clear and unified vision. Historically, banks have undertaken a fragmented approach to AI implementation, particularly with generative AI. A coherent bank-wide strategy is essential to drive efficiency, foster innovation, or enhance risk management. A well-articulated vision should outline specific outcomes, acknowledge risks and costs, align with the bank's mission, and be communicated effectively across stakeholders. A disciplined funding approach is necessary to channel resources towards impactful initiatives, preventing the proliferation of disconnected projects.
Furthermore, banks need to clarify ownership and governance for AI initiatives. Accountability is often dispersed, making it crucial to determine central responsibilities and those of individual business units. Adopting a hub-and-spoke model may provide an effective framework, with a central AI center of excellence overseeing governance and quality across the organization. This entity would manage AI strategy and serve as the operational nucleus for AI integration, including the development of reference standards and shared resources.
Another significant consideration is the build vs. buy decision in AI development. Many banks have employed a hybrid approach, creating proprietary models while purchasing third-party solutions for standard needs. For generative AI, there is a trend towards an assembly approach where banks buy foundational models and build tailored layers around them. This strategy can expedite market readiness and reduce experimentation costs, while transferring some risks to third parties. However, smaller banks may need to rely on this hybrid model due to budget constraints and talent shortages.
Measuring the return on investment (ROI) from AI initiatives is critical yet challenging, as many executives struggle to assess value beyond subjective measures. Advanced ROI measurement is evident in areas like software development productivity, but establishing standard baselines and key performance indicators is often lacking. Without this framework, the perceived benefits may not reliably connect to tangible financial outcomes, creating a credibility gap. Only 4 of 50 banks studied reported measurable ROI from AI use cases by 2025.