BCG: 73% of major banks allow AI agents in production, but only 8% fully control them

Most of the world's systemically important banks now let AI agents run in production for some tasks, but few can fully monitor and control them, according to a benchmark by Boston Consulting Group of 26 global and domestic systemically important banks.

Seventy-three percent of the banks permit agents in production for select use cases, and none allow broad production use. At 88% of banks, agents make up 10% or less of the AI portfolio. Only 8% have a comprehensive, standardised framework for monitoring and controlling agents, and 54% have set restrictions or bans on agentic AI in their risk appetite. Nearly half, 46%, reported an AI-related issue or incident in the past 12 months, and more than a quarter said AI systems were falling short of their promised business case.

Governance basics are mostly in place: 96% manage AI risk through existing risk frameworks and 77% have a dedicated AI policy, but fewer than half keep a central, tool-based AI inventory. BCG says banks with more mature risk management are 2.5 times more likely to be advanced AI adopters. BCG sells AI governance and risk consulting to banks.