Decorrelating the Advice Layer: A Calibrated Fragility Metric and Tested Mitigations for Synchronized Liquidation in Agentic Wealth Management (OpenAlex (recherche mondiale))

Summary

• In a March 2026 survey of 938 U.S. investors, 62% reported using AI tools for investment decisions, 22% used them to assist portfolio decisions, and 24% named "market herding" as a concern.p.2
• In September 2026, one provider accounted for 80.5% of worldwide AI-chatbot usage with a Herfindahl–Hirschman index of H = 0.66.p.2
• The effective number of independent decision-makers in the agentic wealth layer is bounded at about 2.0 at measured consumer concentration.p.1
• Index tail losses are 1.47–1.90 times larger under an agentic architecture than under a periodic robo-advice architecture.p.1
• Dispersing overlay parameters, staging execution, and a participation breaker together change tail liquidation by −63% at no measurable cost to client welfare.p.1
• In historical replays, an agentic layer holding 10% of the market would have deepened the 2020 crash by about 858 basis points.p.1

Key insights

1. View monoculture creates crowding in individual securities rather than market-wide liquidation, multiplying the blast radius of a single compromised provider by about 5.7 relative to a diversified market.p.3
2. Synchronized automation drives market-wide liquidation, with index tail losses rising with the square root of agentic adoption.p.1
3. Provider diversification has no effect on market-wide tail liquidation, changing SLaR99 by +0.2%.p.3
4. A fragility index computed from aggregate flow, public returns, and an impact estimate forecasts next-period tail liquidation with a hold-out R2 of 0.56.p.3