The AI-enabled risk function of the future
68/ 100
Summary
McKinsey argues that after two decades of building risk governance, data and controls, banks face still-mounting demands, and AI now offers a redesign of the risk function, not just automation of single tasks. It describes five shifts — risk work becoming predictive, dynamic, faster and more consistent, with controls embedded in processes and reusable platforms — plus a three-pillar model of a strategic horizontal function, lean AI-enabled domains and a risk-AI centre of competence. Claims include 60 to 70 percent of manual tasks eliminated and 30 to 50 percent productivity gains in early wholesale lending; five roadblocks are flagged, and delay is judged costlier than leadership.
Highlights
With appropriate validation and oversight, AI-enabled workflows can monitor exposures more frequently, interpreting inputs, delivering more timely analysis, compiling evidence, and escalating exceptions to specialists.
We estimate that AI-enabled workflows could eliminate 60 to 70 percent of manual operational tasks, depending on the process and degree of automation.
Early implementations show productivity improvements of 30 to 50 percent across the end-to-end credit life cycle as well as shorter turnaround times and output quality rates of approximately 95 percent.
Contrarian Viewpoints
- The headline productivity and cost figures rest on McKinsey's own client engagements rather than independently audited results, so realised gains at other banks may fall well short of the 60 to 70 percent range.
- The article treats AI's own risks — model error, drift and over-reliance on automated judgement — largely as a governance checklist, which may understate how quickly a flawed workflow can propagate a bad risk decision across portfolios.
Investment Implications
- Long large global banks that industrialise AI risk platforms early, as faster portfolio steering and lower manual risk costs should widen their operating advantage over peers stuck with fragmented point solutions.
- Positive for enterprise AI platform and risk-software vendors selling into bank risk functions, where demand for reusable credit, monitoring and model-lifecycle capabilities is set to grow.