JPMorgan Chase has developed a series of artificial intelligence agents that make investment allocation decisions, and in historical testing the systems outperformed the traditional 60/40 portfolio while producing lower volatility, according to research released by the bank.
The research team, led by Thomas Salopek, found that its best-performing AI model exceeded the annual return of the classic portfolio—comprised of 60% stocks and 40% bonds—by approximately 0.7 percentage point over two decades of back-tested market data. The AI systems also outperformed JPMorgan’s own rules-based investment allocation model on a risk-adjusted basis.
How the AI Agents Work
Rather than simply analyzing market data, the AI agents are designed to make asset allocation decisions.
Using large language models developed by OpenAI and Anthropic, the system evaluates economic conditions and classifies markets into four primary environments: Goldilocks, Reflation, Stagflation and Risk-Off.
The agents then determine how to allocate investments between stocks, bonds and other asset classes based on those changing conditions.
According to JPMorgan, all eight AI agents tested exceeded the performance of both the traditional 60/40 portfolio and the firm’s existing quantitative allocation framework when measured on a risk-adjusted basis.
Back-Tested Results, Not Live Investing
JPMorgan cautioned that the findings are based entirely on historical simulations and should not be interpreted as proof the strategies will outperform in future markets.
The bank noted that back-testing carries well-known limitations, including the risk of overfitting, where models perform exceptionally well using historical data but fail under future market conditions.
Researchers also warned that investment strategies can lose effectiveness as more investors begin using similar approaches.
A New Direction for Wall Street
While investment firms have increasingly used artificial intelligence to summarize research, analyze earnings reports, screen securities and identify investment opportunities, allowing AI to make actual portfolio allocation decisions represents a significant next step.
Because the traditional 60/40 portfolio serves as the foundation for countless retirement accounts, pension funds and institutional investment strategies, even modest improvements in long-term performance could have meaningful implications across trillions of dollars in managed assets.
The Next Phase of AI Investing
Industry analysts say the research highlights how artificial intelligence is evolving from a decision-support tool into a potential decision-maker.
Whether AI can consistently outperform experienced portfolio managers in live markets remains an open question. Real-world investing introduces transaction costs, changing market conditions and investor behavior that cannot be perfectly replicated through historical simulations.
Still, JPMorgan’s research provides another indication that major financial institutions are moving beyond using AI simply to assist investment professionals and are beginning to explore how intelligent agents may eventually participate directly in investment management.
If future live-market performance mirrors even a portion of the historical testing, the technology could reshape portfolio management across the investment industry over the coming years.
JBizNews Desk | New York
© JBizNews.com All Rights Reserved. Reproduction or distribution without written permission is prohibited.
