A version of this article was initially published in the Eurofi magazine.
Over the past two years, financial institutions have made remarkable progress with artificial intelligence. Thousands of use cases have been identified. AI assistants are moving into customer service, software engineering, research, and operations, where many banks are already seeing measurable productivity gains. Large institutions have established AI organizations, governance structures, and growing portfolios of production deployments.Despite that progress, many banks remain focused on the mechanics of adoption rather than what happens next.
Scaling AI is not simply a larger deployment exercise. Moving from 10 use cases to 100 or from one assistant to thousands of employees will not, by itself, produce a different institution. The larger shift comes when AI changes how banks operates, how decisions are made, and where competitive advantage comes from. Most investment has focused on improving existing operating models. Yet the greater opportunity lies in redesigning them.
Scaling AI requires redesigning banking operating models
In conversations with banking executives over the past year, we've noticed a shift. The discussion has moved beyond copilots and toward questions like whether junior analyst programs still make sense and how relationship managers will supervise AI-generated advice.
A year ago, many executives focused on where AI could improve productivity. Today they are thinking about how AI changes organizational design, talent development, and competitive positioning. The discussion has become less about deploying tools and more about redesigning institutions around a different economic model.
These leaders see that AI offers much more than improved efficiency across existing processes; it can help a bank redesign client journeys, operations, and control functions around AI to improve its entire cost structure and operating model. That allows people to spend less time on routine execution and more time exercising judgment, managing client relationships, and supervising increasingly capable AI systems.
Building trust as AI reshapes banking decisions and customer relationships
Technology alone cannot produce that outcome. The main obstacle is trust. Customers have always expected banks to make decisions that are understandable, reliable, and accountable. That expectation will only become more important as AI systems begin recommending products, comparing providers, and, in some cases, acting on behalf of customers. Increasingly, those decisions must also be legible to AI agents acting on behalf of customers. Explainability and transparency are becoming commercial capabilities, alongside regulatory obligations, because both people and machines will influence how financial products are selected and recommended.
The customer relationship has been changing. Digital banking strengthened direct connections by making the bank's application the primary interface. AI assistants may weaken that advantage. If customers increasingly rely on intelligent agents to search, compare, and execute financial decisions, advantage shifts toward institutions those agents judge to be trustworthy and reliable.
The workforce presents another challenge. Banking has traditionally developed expertise through apprenticeship, with junior employees learning through repetitive analytical and operational work before taking on greater responsibility. AI is beginning to absorb many of those early-career tasks, forcing banks to find new ways to develop judgment, client skills, and institutional knowledge when much of that training ground has changed. That may require rethinking career paths, performance management, and even what qualifies someone to lead teams whose work is increasingly performed alongside AI.
Governance must evolve as AI scales across the enterprise
Governance must evolve as well. Banks have spent decades refining governance for people, products, and financial risk. AI introduces another operating layer that requires equally clear ownership and accountability. Frameworks designed for dozens of models will not be sufficient when banks manage thousands of AI agents across business functions. Ownership, permissions, accountability, monitoring, and lifecycle management become operating disciplines rather than compliance exercises.
Investment may ultimately prove to be the greatest constraint. Building enterprise AI capabilities requires far more than foundation models. Data infrastructure, technology platforms, governance, process redesign, talent, and organizational change all compete for capital. Institutions with stronger earnings can reinvest more aggressively, accelerating improvements that weaker competitors struggle to match. That creates a practical advantage: Well-capitalized banks can test, learn, and redesign faster, while others remain trapped in narrower, incremental programs.
The industry has spent the past two years learning where AI can improve productivity. The next phase will depend on whether banks are willing to rethink how the institution itself is designed. Those that make that transition are likely to build advantages that could shape the competitive landscape for years to come.