How Gen AI Reinvents Banking Operations to Unleashing Next-Level Performance

Scaling Generative AI: Overcoming the Pilot Plateau

Launching small generative AI initiatives can happen quickly. The real challenge is making them stick across an organization to generate meaningful impact. Many banks see promise in Gen AI’s capacity to streamline operations, enhance customer engagement, and build entirely new products, but they also acknowledge the risks. Two-thirds of senior digital and analytics leaders at a recent Santiago & Company forum on gen AI reported that this technology will fundamentally alter how they do business. Despite the enthusiasm, banks must consider where and how to deploy gen AI most effectively—and how to ensure lasting adoption that goes beyond proof-of-concept pilots.

According to MGI and Santiago & Company research, gen AI could add the equivalent of $2.6 trillion to $4.4 trillion annually across the many industries it studied. Among sectors, banking may see one of the greatest gains: $200 billion to $340 billion annually, or about 9 to 15 percent of operating profits, mainly through productivity increases. Those gains span nearly every banking segment and function. While initial efforts often emphasize improving productivity in response to economic pressures, gen AI could also reshape how employees work, how customers interact with banks, and the types of business models that emerge.

Building AI Readiness and Addressing Workforce Needs

At the same time, the technology raises the bar for aligning business, technology, and analytics. While many institutions felt they had bridged the “business-technology” gap through agile methods, cloud adoption, and changes in product operating models, gen AI has revealed that data and analytics now form a critical third link. Business leaders must work more closely with analytics teams, especially as gen AI threads its way into more processes—from customer interactions to back-end systems—and demands deeper synchronization on multiple priorities.

Another complication is the technology’s rapid uptake. Smartphones, for instance, took years to move banking to its current digital state. In contrast, gen AI solutions are being adopted at remarkable speed. Goldman Sachs has used AI to automate certain testing activities previously done by hand, and Citigroup recently assessed the impact of new US capital rules with gen AI. For institutions unaccustomed to moving this quickly, keeping pace may strain their operating models. A final challenge revolves around the workforce. Some banks, particularly those with a strong base of quantitative and AI-savvy employees, are better positioned to adapt; they can train existing staff in emerging skills such as prompt engineering. Others must embark on extensive hiring campaigns to fill capability gaps. Both approaches require sustained investment in people, processes, and technology.

Successfully Scaling  Gen AI Across Seven Dimensions

Although scaling gen AI poses unique challenges—such as model tuning and ensuring data quality—getting started can be simpler than a traditional AI endeavor of similar scope. Capable teams often launch valuable use cases in days or weeks. Yet moving beyond pilots to achieve sustained value calls for robust capabilities in seven interconnected areas: strategic road map, talent, operating model, technology, data, risk and controls, and adoption and change management. When banks align effectively along these dimensions, they build a solid foundation that supports gen AI’s rapid deployment and helps each new solution take root in daily workflows.

1. Strategic road map

Banks that have achieved early success typically begin by visualizing where gen AI, AI, and advanced analytics can best fit their operations. They weigh everything from transformative overhauls of the business model to smaller, highly targeted initiatives aimed at boosting productivity. For example, a wealth management firm recognized that gen AI could eventually revolutionize how advisors serve clients and how the broader ecosystem—including partners, platforms, and third-party providers—operates. Leaders there adopt a flexible approach to deciding where to invest, preparing themselves for a future in which gen AI might open several different pathways at once.

This type of big-picture thinking hinges on strong alignment among senior leaders, who set clear targets and hold business units accountable for results. A strategic road map also details the domains (or functions) with the greatest potential to yield business value, explains how to build capabilities such as agile operating structures and robust technology, and clarifies the investment required in partnerships or third-party solutions. When leaders unify around a coherent, enterprise-wide perspective, banks can eliminate duplication of effort and deploy their resources wisely.

2. Talent

The rapid rise of gen AI has left little time for banks to assess the implications for their people fully. Leaders themselves must gain hands-on experience with this evolving technology so they can communicate clearly how it connects to the bank’s strategic goals. Targeted executive education can accelerate their comfort with key concepts and generate the necessary energy to encourage broad adoption.

Building credibility among employees may begin with two or three high-profile initiatives that illustrate Gen AI’s potential. These “lighthouse” projects help employees see how gen AI automates certain tasks, and they give leaders a chance to explain how the technology can enhance—rather than replace—the roles people perform. Naturally, questions about job security arise whenever automation is discussed. Banks that address these concerns directly through transparent communication are more likely to earn employee support and encourage collaboration.

The spread of gen AI also creates new job roles. Prompt engineering and model fine-tuning were seldom mentioned in most banks’ talent strategies before this year. No single institution is likely to have all the expertise it needs at the outset, so each one must decide on the right mix of internal training and external hiring. Such investments should be part of a broader, long-term approach, reflecting the reality that gen AI will keep evolving, and use cases that prove vital next year might not even be on the radar now.

3. Operating model

Institutions often talk about designing a “new operating model” whenever a technology trend emerges, though experience suggests it’s better to refine existing ways of working so they can adapt to breakthroughs like gen AI. A cross-functional approach, in which product teams collaborate with business leaders, technology specialists, risk managers, and data professionals, is crucial. The most successful initiatives foster a transparent process for allocating budgets, approving projects, and championing new features.

Given gen AI’s novelty, many banks have opted for a more centralized framework, where a single function sets standards, steers research manages vendor partnerships, and coordinates large-scale use cases. Santiago & Company analysis of US and European banks shows that over half favor a more centralized organization of gen AI, even if other analytics functions remain relatively decentralized. This tactic can accelerate learning, streamline governance, and ensure consistency in risk controls. Yet centralized teams must still engage with business leaders early and often, integrating valuable frontline perspectives into model design and deployment. Over time, banks might pivot to a more federated approach in which select domains gain additional autonomy, but the most successful institutions begin with solid central guidance.

4. Technology

In deciding how to use gen AI, banks often debate whether to build their own solutions or rely on partnerships and ready-made tools. The line between internal development and third-party vendors can be fluid since open-source models evolve quickly and may soon become widely accessible. Banks that take a thoughtful look at which capabilities indeed confer a strategic advantage are more likely to scale effectively.

Still, working with new AI models requires a level of trust that some banks have never granted external providers. Leaders must understand how gen AI’s inherent risks—such as uncertainty about model outputs—fit within existing risk tolerances. Some banks might deliberately limit the applications they develop to areas that remain below certain thresholds of regulatory or reputational exposure.

Integration is another hurdle. If a bank pursues multiple gen AI use cases, each needs to mesh with existing data flows, enterprise applications, and legacy systems. A fully realized gen AI platform often includes context management, caching, policy management, a model hub, prompt libraries, model-operations capabilities, and other elements that must align cohesively. Achieving that alignment requires a holistic approach that recognizes gen AI as part of a broader technology ecosystem, not a disconnected add-on.

5. Data

Gen AI’s appetite for unstructured data introduces another layer of complexity. Many banks have moved some or all of their data to the cloud or to specialized third-party providers, which can create new constraints or open new possibilities. They also face ongoing challenges in operationalizing unstructured data—such as text from customer messages or social media—that can supercharge gen AI if used correctly, but often lies dormant because the bank lacks the right tools.

Gen AI itself offers promising ways to tackle these issues. Its natural language capabilities can parse enormous volumes of unstructured data and prompt employees to interact with information more intuitively. In customer service settings, for example, frontline agents might receive relevant conversation snippets or credit details in near real-time. These emerging solutions can also help banks build exemplary data architecture. Tools like vector databases and sophisticated pre- and post-processing can help embed gen AI throughout the organization. Data quality remains a top priority, though, because unstructured data at scale can easily harbor inaccuracies. Banks that succeed typically combine manual expertise with automated checks, ensuring data is trustworthy and up to date.

6. Risk and controls

Gen AI expands the risk landscape as it expands opportunities. Because it can generate outputs that are factually incorrect, biased, or at odds with a bank’s own guidelines, banks must strengthen their existing risk frameworks with additional guardrails. Most institutions are still discovering how to adapt risk and model governance to generate AI’s faster, less predictable environment. They need a plan for responsible gen AI from the outset.

Many banks rely on human experts to verify model outputs, but such reviews become laborious when scaling to dozens or hundreds of use cases. Some have responded by creating automation, validation methodologies, and playbooks that anticipate “hallucinations” or inaccurate outputs. Simple configuration steps—like adjusting a model’s temperature settings—can help gen AI deliver more consistent and reliable answers, while automated content filters can screen for harmful or misleading language. Risk leadership should also pay close attention to evolving guidelines and regulations for gen AI, anticipating that requirements for model explainability and fairness may increase over time.

7. Adoption and change management

Even the most potent gen AI tool has little impact if employees or customers are reluctant to use it. A bank might create an innovative AI platform that falls flat because the workforce doesn’t trust the results or find the interface intuitive. Some executives may also hesitate to champion an initiative if they’re unclear about who “owns” it or how it fits into the product portfolio.

A user-centric mindset can mitigate these concerns. Leaders who begin with frontline experience and design processes around real-world needs typically see better adoption. They can create AI agents that respond to dynamic user feedback, improving these tools over time. Banks that support this transition with a thorough change management plan—incorporating training, incentives, and open communication—stand the best chance of unlocking lasting value. The plan should acknowledge employees’ anxieties about job displacement while showing them how gen AI frees up time for more strategic, creative work.

In addition, leaders and influencers must visibly model the shift. If senior executives use and endorse gen AI capabilities, they signal the bank’s commitment to innovation and help normalize the new technology among employees. Clear priorities, practical guidelines, and user-friendly platforms also encourage staff to experiment and discover how gen AI can simplify daily tasks or open up new possibilities.

Gen AI undoubtedly holds vast potential for banks, promising to increase productivity, enhance customer engagement, and spark creative reinventions of core processes. Success, however, depends on the ability to embed these capabilities within the larger organization. That path calls for a synchronized approach across strategy, talent, operations, technology, data, risk, and change management. Banks that make consistent investments in these seven dimensions and remain vigilant about gen AI’s evolving risks will likely be the ones that translate early pilots into enduring, transformative value. Over time, as more institutions learn to navigate gen AI’s challenges and fully absorb its potential, the banking industry may witness innovations that we have only begun to imagine.