The Bridge Ep. 16: Can AI See What Investors Miss?

Summary

In this episode, iCapital’s Sonali Basak speaks with Greg Bond, CIO of Man Group, about why quant hedge funds have performed well and where risks may be building underneath the surface. Bond says the current backdrop of higher rates, dispersion, and periodic market shocks has helped create opportunity, but investors need to separate true alpha from hidden beta, especially in crowded or highly levered multi-strategy environments. He argues that liquidity and horizon diversification are the main defenses against sharp drawdowns, and that many of today’s market interconnections are obscured by capital-efficient platform structures. On AI, Bond says the biggest impact is in the research process: creating “digital researchers,” encoding firm-specific judgment into harnesses or skill files, and using simulation to stress test portfolios across many synthetic market worlds. His core message is that AI can expand what investors can see and test, but it cannot replace experience, context, or rigor.

Notes

• Man Group is framed as the world's largest publicly listed hedge fund firm, with almost $230 billion in AUM.
• Greg Bond is the CIO and also manages Man's flagship Malta strategy portfolio.
• Quant strategy performance has been strong because rates moved back toward more normal levels.
• Dispersion trading benefited from more shocks in the system and more cross-sectional opportunity.
• Bond says investors should dig into strategy drivers and not assume performance equals true diversification.
• He warns that some apparently strong returns may actually be hidden beta rather than alpha.
• Post-COVID, managers kept gross exposure and leverage elevated to absorb small dislocations.
• Bond describes these recurring small shocks as 'quant quakes' and 'discretionary quakes.'
• Liquidity is a first-line defense because liquid strategies can adjust to real-time information.
• Trend strategies may help in sustained drawdowns.
• Quant has evolved from a couple of models to hundreds of models and techniques.
• Manager sets have become more diversified than in the early 2000s.
• He says liquidity risk management is better now than during earlier small-cap-heavy periods.
• Some signals have been pushed to longer horizons, including 'long term alpha signals.'
• Bond distinguishes investing-style signals from shorter-term trading signals.
• Competition is increasing as higher-frequency and mid-frequency players converge.
• Momentum works over long periods but has to be defined carefully in practice.
• Bond prefers fundamentally driven momentum such as earnings revisions.
• He points to supply-chain and ecosystem effects as part of fundamentally driven momentum.
• Retail participation is rising in smaller names with distinct liquidity profiles.
• Meme-stock behavior is cited as an example of retail-driven dynamics.
• Multi-strat and other crowded trades can create contagion across portfolios.
• Index rebalancing trades can spill into traditional strategies.
• Calendar effects can amplify pressure when many participants face the same schedule.
• Deleveraging by systematic and multi-strat players can converge into a selloff.
• Market structure may look diversified, but hidden correlations can surface under stress.
• Multi-strat platforms are capital-efficient because one manager can see and adjust real-time risk across many strategies.
• Platform structures can hide interdependencies even when they appear diversified.
• Bond says his bigger worry is confidence in data and information, not a pure market collapse.
• Trust in institutions and data quality is part of market stability.
• AI has been most valuable in the research process, not just in building models.
• Bond imagines creating the equivalent of a 'digital researcher.'
• AI can add more researchers effectively by widening the research workforce.
• The key AI moat is increasingly the harness or skill file around the model.
• Firms need to teach AI their institutional DNA, including standards for good research.
• Bond says AI should be used with multiple reviewers and active checking.
• He warns that AI can be fast but lacks context and experience.
• Hallucination risk is a practical concern in AI-assisted research.
• He suggests using many AI reviewers, even eight or nine or ten, to compare outputs.
• A promising AI use case is generating simulated market worlds for stress testing.
• He wants to test models against many synthetic environments, not just one historical crisis.
• Stress testing should ask not just the probability of a meltdown, but why it would happen.
• AI may improve causality analysis in risk management by helping answer 'why.'
• The episode’s broader takeaway is that AI expands research capacity, but judgment still matters.
• Investors should focus on liquidity, diversification, and hidden correlations rather than headline returns.