Why Agentic AI Will Transform Banking Gradually

Across the financial services industry, AI is no longer a question of “if,” but “how fast and how far.” As discussed in this interview, every institution — regardless of size — is actively exploring how artificial intelligence can reshape its operations, products, and decision-making. But the path forward will not look the same for everyone.

Smaller banks, in particular, are well positioned to move faster. With lighter regulatory burdens and simpler operating models, they can deploy AI in areas where the impact is immediate and measurable. Faster credit approval cycles, accelerated underwriting, enhanced fraud detection, and advanced analytics are all within reach. These institutions can also move toward bespoke, tailored solutions, using AI to personalise offerings in ways that larger organisations often struggle to execute.

By contrast, wholesale and investment banks face far greater complexity. Their focus is increasingly on managed services and off-the-shelf AI solutions that can simplify sprawling processes and drive operational efficiency at scale. For them, AI is less about radical reinvention and more about rationalisation — reducing friction, lowering costs, and improving consistency across vast organisations.

Yet the common thread is transformation. AI is not replacing human expertise; it is augmenting it. The real opportunity lies in combining human judgment with machine intelligence to create better outcomes — faster decisions, deeper insights, and more resilient systems.

The conversation then turns to agentic AI, widely seen as the next frontier. While some predict fully autonomous organisations within a few years, the reality is likely to be more measured. The interview suggests that 2025 marks the beginning of the agentic AI era, not its endpoint. Institutions are already experimenting with agent-based systems, particularly in internal network units, but not yet in customer-facing roles.

Adoption will follow a staged approach. Banks will begin with low-risk, high–data-volume agents — level one use cases — before progressing through increasingly complex scenarios. High-risk applications, especially those affecting customers directly, will continue to require human oversight. This progression is expected to unfold over the next two to three years, with the five-year horizon remaining highly uncertain.

Beyond the enterprise, deeper questions loom. As organisations become more dependent on AI models, societal and geopolitical implications emerge. What happens if entire sectors rely on a small number of AI providers? How does this concentration of capability affect national resilience, competition, and autonomy? These are the quieter, less visible debates shaping the future — but they may prove the most consequential.

The industry is not just changing its tools. It is changing its future business model with AI, step by step.