AI to Directly Perform Financial Tasks... “Designing Authority and Accountability Comes Before Automation”

[Financial News] As artificial intelligence (AI) moves beyond assisting financial institutions with their operations and enters the era of 'AI agents' that make decisions and take action directly, the direction of process innovation (PI) in the financial sector is also changing. The key challenge is no longer simply how much work can be automated, but how much authority should be given to AI, when human intervention should occur, and who will bear ultimate responsibility.
In a report published on the 6th, Deloitte Korea Group said, "PI at financial institutions needs to expand from improving efficiency in people-centered operations to building a 'Target Operating Model' in which people and AI work together."
While conventional PI focused on reducing the number of work steps and saving time and labor, the AI-agent era requires the roles, authority, responsibilities, and control systems of people and AI to be designed simultaneously.
Deloitte proposed four mechanisms for operating AI agents at financial institutions: authority controls, decision controls, audit-trail controls, and exception controls.
The key is 'exception handling.' If financial institutions do not establish when AI should stop making decisions and hand cases over to people in situations it struggles to handle—such as high-value transactions, conflicting data, or abnormal signs—operational risks could actually increase even as the automation rate rises.
Meanwhile, global financial institutions are already restructuring their AI operations in this manner. Allianz operates 'Project Nemo' in Australia to process the surge in small insurance claims following natural disasters. The corporate and investment banking (CIB) payments division of JPMorgan Chase also used AI and machine learning to reduce by more than 50% the number of exceptions requiring manual processing and cut exception-handling time by about 75%.
Byun Sung-wook, insurance industry leader at Deloitte Consulting's Financial Consulting practice, said, "In the era of AI agents, what matters is not how many tasks have been automated, but how the roles and authority of people and AI are designed and who bears ultimate responsibility in exceptional situations." He added, "In the financial sector, performance and risk depend on how smoothly people can take over and resolve exceptions that AI has difficulty handling."
[email protected] Kim Hyun-jung Reporter