NY Fed Risk Chief: AI Efficiency Gains Could Hollow Out Bank Leadership Pipelines
The Federal Reserve Bank of New York's chief risk officer warned that automating the entry-level work that trains future bank leaders could hollow out leadership pipelines and institutional memory within five to seven years.
Speaking at the Risk Live North America conference, Mihaela Nistor, Chief Risk Officer and Head of the Risk Group at the Federal Reserve Bank of New York, said the gap between how fast institutions can deploy artificial intelligence and how fast they adapt organizationally is itself an operational risk that risk managers must own. Ms. Nistor said large, systemically important institutions absorb technology unevenly: a business unit may automate junior analytical work before governance structures catch up, or deploy autonomous agents before resilience teams design a fallback. That asynchrony among technology, controls and workforce capability, she said, is where institutional fragility emerges.
Ms. Nistor distinguished efficiency (doing normal work with fewer resources) from resilience (the capacity to adapt to failures institutions did not anticipate). She cautioned that optimizing away entry-level work could silently shift risk into the future without triggering any near-term control failure.
She urged risk professionals to stress-test the institution, not just the technology. She named three medium-term risks that belong permanently on risk radars: (i) erosion of human capital and leadership benches; (ii) AI-embedded third-party dependencies that fall outside traditional vendor-risk frameworks; and (iii) geographic and skills mismatches.
She also flagged three longer-term "acceleration pathways" to treat as strategic stress scenarios: (i) the convergence of AI and robotics, which could erode the protection from automation that physical work has long enjoyed; (ii) autonomous agents changing the ratio of human supervision to output; and (iii) concentration of compute, energy and semiconductor capacity, which could create new asymmetric risks.
Ms. Nistor closed by asking risk leaders to consider not only what happens when AI fails, but what happens to institutions when it succeeds.