As banks adopt agentic AI to boost efficiency, leaders grapple with the challenge of balancing autonomous decision-making with organisational oversight, amid rising concerns over control and responsible deployment.
Bank leaders are increasingly convinced that the hardest part of artificial intelligence in banking is no longer building the tools but deciding how much independence to give them. At FIBAC 2026, Ashok Chandra, managing director and chief executive of Punjab National Bank, said lenders need to be far more cautious as they move from traditional machine learning and generative AI towards agentic AI, where software can act with less direct human prompting. He warned that if AI agents are not properly controlled, they can start behaving differently at an organisation-wide level.
That concern is not limited to one bank. DBS Bank has already embedded AI agents in its corporate lending process, but Computer Weekly reported that it is stopping short of letting them operate on their own because oversight systems are not yet strong enough. Nimish Panchmatia, DBS’s chief data and transformation officer, said the technology is advancing faster than the controls around it, leaving human supervision necessary for responsible deployment.
The wider banking industry is facing the same tension between speed and control. McKinsey has warned that generative AI agents could sharpen efficiency in retail banking while also weakening direct customer relationships if banks automate too aggressively. EY has argued that agentic AI is reshaping competition and operating models across banking, but says firms will need to redesign those models carefully if they want to preserve control over outcomes while still capturing the benefits of automation.
Practical barriers remain a major brake on scale-up. Backbase has pointed to fragmented data, heavy compliance demands, legacy systems and complex multi-entity structures as recurring obstacles in commercial banking. LatentBridge has likewise highlighted problems such as broken context, mismatched workflows and governance bottlenecks. Unblu says the gap between experimentation and deployment is already visible, with many financial services firms running pilots but far fewer moving AI agents into full-scale use.
Disclaimer: This article is intended to inform and educate, not to recommend or endorse any financial product, investment or strategy. Please consider your own financial circumstances and seek professional advice where appropriate before making financial decisions.
