Every so often, humanity engineers a technological breakthrough so transformative that it reshapes the course of civilization. For example, the printing press, introduced around 1440, did more than make books affordable — it democratized knowledge, fueled the 16th-century European Reformation, ignited the scientific revolution, and laid the groundwork for modern democracy. In the realm of technological and economic revolutions, other milestones stand out as well: the industrial revolution, electrification, the internet, and now, artificial intelligence (AI).The Promise of AIAI has the potential to transform the banking industry by enabling both significant revenue growth and cost reduction, including through better customer targeting, personalized product offerings, automation, streamlined operations, and more. In its recent analysis of how AI is reshaping the industry, PWC suggests financial institutions that fully embrace AI could see substantial efficiency gains, with up to 15% improvements in the efficiency ratio by cutting cycle times, reducing manual workloads, and driving smarter, data-driven decision making. AI also offers the promise of reimagining customer engagement — from tailored financial advice to proactive support — and transforming banking from transactional interactions to more contextual services that anticipate customer needs, deepen relationships, and unlock new growth opportunities. According to a KPMG survey, US banking executives have strong confidence in AI’s potential to transform their business. More than four in five believe that institutions that embrace AI will develop a competitive edge over those who do not. This belief is reflected in their budgets: 82% plan to increase their global spend on AI, with 62% targeting up to 20% increases and 38% planning increases of more than 20%.Mixed ROI The industry’s track record, however, is mixed. A recent analysis of AI outcomes in US banking conducted by Evident found that 70% of AI initiatives have little to show in terms of return on investment (ROI) because their efforts remain fragmented and experimental. The analysis also found a growing divide between institutions that treat AI as a strategic capability and those that pursue pilots without scale or accountability. Across several studies that look at the impact of AI, three common themes emerge.
- Incremental, bottom-up AI is of limited value. Many organizations fall into the trap of pursuing superficial, task-level AI initiatives driven by individual employees. Rather than creating meaningful value, these implementations often generate additional work downstream, requiring review, correction, and rework, which can erode productivity and obscure the true potential of AI.
- AI implementations should be paired with process redesign. AI delivers significant value only when it is paired with intentional process redesign, not when it is layered onto existing ways of working. Organizations must rethink workflows end-to-end in terms of outcomes, not tools — clarifying decision rights, redefining roles, and embedding quality controls so humans and AI can each do what they do best.
- Organizational barriers matter more than technology. Common challenges include a lack of clear ownership and accountability, weak or fragmented data foundations, and risk-averse cultures that slow decision making. These issues are often compounded by misaligned incentives and insufficient employee buy-in, which can limit adoption and prevent initiatives from delivering meaningful impact.
Lessons for CFIsAs community financial institutions (CFIs) consider their approach to AI, there is much they can learn from the experience of early adopters in ensuring an effective implementation and strong ROI. Here are four practical tips:
- Adopt a strategic top-down approach. Prioritize high-value use cases that align with core business objectives — such as improving customer engagement, reducing processing times, or enhancing risk management — while redesigning workflows to optimize the balance between AI and human intervention. Bear in mind that routine, repetitive processes — such as loan processing, fraud detection, document review, or cash-flow forecasting — tend to deliver more reliable ROI than flashy generative tasks.
- Focus on organizational readiness. To integrate AI effectively, CFIs need to hire and train the right talent and overcome cultural and behavioral barriers by positioning AI as a tool that allows employees to spend more time on strategic and creative tasks. Institutions also need to establish a strong governance structure with clear roles and ownership, centralized standards, alignment across teams, and incentives that are linked to desired outcomes.
- Ensure data readiness. AI systems rely on high-quality, structured, and accessible data to generate reliable insights and automate decisions. Without clean, complete, and well-governed data, AI outputs can be inaccurate, inconsistent, or require significant manual correction, which reduces productivity and erodes trust. CFIs should invest in standardizing and cleaning core data sources and establishing clear data ownership and governance, including common definitions, quality controls, and access protocols.
- Monitor continuously. Many AI benefits often appear in indirect metrics like faster decision making, fewer errors, improved customer satisfaction, and reduced manual effort, rather than immediate cost savings. CFIs should carefully choose a set of key performance indicators and regularly track both quantitative and qualitative gains to capture true value and guide ongoing optimization.
AI’s promise is real, but its returns depend less on the technology itself and more on how organizations deploy it. Incremental use cases, poor governance, weak measurement, and cultural resistance are holding many institutions back. The strongest ROI comes from adopting a strategic approach, reengineering core processes, focusing on organizational and data readiness, and measuring progress. CFIs that treat AI as a transformation effort — not a productivity tool — are far more likely to see long-lasting, compounding value.
