Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Ai Investments And Governance topic
No spam. Unsubscribe anytime.
Asset managers tell CalPERS AI adoption will reshape portfolios, raise data‑center, governance and supply‑chain risks
Summary
Speakers from Man Group and Goldman Sachs told CalPERS trustees AI adoption is fast and pervasive; they urged layered governance, robust testing to avoid "cognitive surrender," and attention to token costs, data‑center siting and subsea‑cable vulnerabilities.
Robin Grew, CEO of Man Group, and delegates from the Goldman Sachs Global Institute briefed trustees on the investment and operational consequences of rapid AI adoption.
Grew emphasized that AI is now a ubiquitous tool across research and operations, and she urged trustees to guard against what she called "cognitive surrender"—an overreliance on models without human review. "AI is not going to replace humans, but humans who don't use AI will probably be replaced," Grew said, and she recommended layered agent checks, backtesting and active human oversight.
She described high internal adoption rates at Man Group, where operations and research teams use AI to accelerate work, and said managers should demand transparent governance from their portfolio companies: prompt‑engineering standards, human review layers, and testing that tracks both failures and successes.
Goldman Sachs' George Lee and Jared Cohen discussed AI's geopolitical and infrastructure implications: token costs for agentic models are rising, a shortage of compute and high‑density data centers is constraining adoption, and critical old‑technology infrastructure—subsea cables and land‑based power—remains essential and vulnerable. Cohen described the token economy and the risk that only countries with capital, compute and energy can sustain large‑scale AI deployments.
Both managers flagged environmental and security issues: large AI workloads demand significant energy and water for data centers, and recent regional security events have shown those facilities can be targets. They encouraged trustees to evaluate managers and vendors on governance, testing and resilience to supply‑chain and geopolitical shocks.
Trustees asked detailed questions about model testing, tracking failures, adopting AI in a public‑sector context and how smaller managers and portfolio companies can be evaluated for responsible deployment. Grew recommended asking managers to demonstrate governance, back‑testing and human review and to share implementation plans.
