Agentic AI in Banking: Risks & Rewards for Financial Services

Banks are rapidly moving into a new phase of AI deployment, using autonomous software agents that can make decisions and execute workflows with limited human input. These agentic AI systems are being experimented with across financial services for tasks like fraud investigations, compliance monitoring, and customer interactions. What's driving this push is the need to cut operational costs and increase productivity. However, there are growing concerns among regulators and testing teams about the long-term risks. A new research paper warns that while these systems offer efficiency gains in structured workflows, they could introduce new systemic and operational risks in more ambiguous environments. The paper, "The Rise of Agentic AI in Banking," argues that banks need more granular quality assurance and governance frameworks for these autonomous agents. It highlights that agentic AI systems are capable of proactive goal-setting and autonomous execution without constant human prompting. By early 2026, major institutions like Goldman Sachs and JPMorgan Chase had already integrated autonomous AI into key workflows. The research identifies significant efficiency gains, including reported net cost reductions of up to 20% and projections of £100 million in annual value from automating fraud investigations. The bottom line: While agentic AI offers substantial benefits, understanding and mitigating its risks is crucial for the financial industry.