After years of experimentation, US banks have largely answered a critical question: Does AI work? The answer is yes, and increasingly so. As artificial intelligence moves beyond pilots and becomes embedded across core banking operations, the US AI in banking market is projected to grow from $14.5bn this year to $31.2bn in 2031. The drivers are clear and pragmatic: cost optimisation, cybersecurity and customer experience. Banks are deploying AI‑powered solutions to automate operations, detect and prevent fraud, and deliver real‑time, personalised customer interactions.
What distinguishes the current phase is intent. AI is no longer being explored as a promising technology. It is being adopted as a business capability. US banks, in particular, are shifting focus from experimentation to value realisation, integrating AI where it can demonstrably improve efficiency, resilience and growth.
US banks are at the forefront of a global trend
While US banks are leading this transition, AI maturity is advancing across the global banking industry. The latest Infosys Bank Tech Index research reveals that banks worldwide are making more deliberate technology choices, with greater clarity on where AI can deliver meaningful impact. Institutions are launching more initiatives, but doing so selectively, prioritising use cases with quantifiable and sustainable returns.
Go deeper with GlobalData
Access deeper industry intelligence
Experience unmatched clarity with a single platform that combines unique data, AI, and human expertise.
This shift in approach is visible in changing strategic priorities. In 2023, cost reduction overwhelmingly dominated AI investment decisions. Today, while cost remains the top driver, it is closely followed by innovation and business growth. Another hallmark of maturity is sharper scrutiny of initiatives. Banks are no longer experimenting indiscriminately and are increasingly exiting projects with weak ROI. The previous edition of Infosys Bank Tech Index found that participating banks had canceled 6,100 AI initiatives before deployment. That figure rose 33% to 8,100 in the latest study. At the same time, cancellations after deployment declined from 3,700 to 3,400, indicating stronger upfront discipline. As a result, approximately 59% of deployed AI initiatives among participating banks are now generating measurable business value.
Business value and cost optimisation go hand-in-hand
Customer service has emerged as the most valuable AI use case for banks. Intelligent virtual assistants, automated service workflows and personalised engagement are reducing cost to serve while improving satisfaction among digital customers. At Citizens Bank, for example, an AI‑powered virtual assistant providing personalised support for routine banking tasks reduced mobile‑app‑driven calls to the contact center by approximately 44%. At Danske Bank, an AI assistant for financial advisers cut average call time from six minutes to under one minute, enabling faster and more accurate responses.
Sales and marketing, cybersecurity and business operations are the other areas where banks are realising significant value from AI. Not coincidentally, these functions also offer some of the greatest opportunities for cost reduction. One notable exception is software engineering, which ranks lower in near‑term business value. The reason is adoption rather than potential: AI‑assisted coding and testing are expected to reduce the unit cost of software development over time. However, because many banks have yet to adopt these tools at scale, their full impact is not yet reflected in value realisation metrics.
Scaling remains the hardest challenge
Despite growing AI maturity, scaling AI across the enterprise remains difficult. Legacy systems continue to constrain model development and deployment, while poor data quality, inadequate risk controls, unclear business cases and high operating costs are expected to cause at least 30% of GenAI initiatives to fail after proof of concept. Many pilots stall before production due to integration complexity, monitoring overhead and change‑management costs.
Banks’ regulatory, risk management and governance obligations further raise the bar. Every model must be rigorously tested and validated before deployment. Scaling AI therefore requires close collaboration among AI engineers, data scientists, business leaders and compliance teams. Yet many banks still treat AI as a technology function rather than an enterprise operating‑model shift. Talent shortages, limited AI fluency, particularly at senior levels and cultural resistance continue to slow adoption.
US banks should address these barriers with urgency. AI is already a powerful differentiator in banking, but its benefits accrue unevenly. Pilots and isolated deployments may demonstrate promise, yet meaningful returns emerge only when AI is scaled across the enterprise. Institutions that fail to move beyond experimentation risk leaving significant cost, productivity and profitability gains unrealised.
Evidence suggests the gap between leaders and laggards is widening. The ten most AI‑mature banks are advancing far faster than their peers and capturing measurable business value through early, disciplined investments in AI while US banks have made real progress in building AI capability. The next phase, and the defining one, will be converting that maturity into enterprise‑wide execution at scale.
Ajay Bhandari is SVP and Regional Head of US Banking at Infosys
