McKinsey: agentic AI could lift bank compliance productivity up to 20x, far beyond earlier AI tools

McKinsey research highlighted by identity-platform Duna finds that banks detect only about 2% of global financial crime despite spending hundreds of billions of dollars annually, with up to 20% of headcount tied up in KYC/AML work. Earlier analytical and generative AI tools delivered modest, roughly 20% productivity gains by assisting human investigators. Agentic AI, where a “workforce” of AI agents handles onboarding, monitoring, and case review end-to-end with humans supervising by exception, can push productivity gains to 200%–2,000%, McKinsey estimates, since one compliance officer can typically oversee 20 or more agents.

The productivity ceiling for compliance AI just moved from incremental to transformational, but only with end-to-end redesign, not bolt-on tools. Banks should pilot agentic workflows in KYC/AML now rather than extend existing assistive AI.