JPMorgan AI Agents Beat Traditional Portfolio in Asset-Allocation Tests

Wall Street investment firms have long used the “60/40” stock-bond allocation as a classic benchmark for diversification. Advances in AI are now prompting the financial industry to explore the technology’s potential in asset allocation. A recent study by JPMorgan’s cross-asset strategy team used multiple AI agents for the first time to backtest dynamic portfolio adjustments, pointing to a new phase in investment automation.

JPMorgan said in a July 9, 2026, report that eight AI agents developed using OpenAI and Anthropic models all outperformed a 60/40 portfolio in historical backtests covering the past 20 years. The top-performing system delivered an annualized return 0.7 percentage point higher than the benchmark, with lower annualized volatility. Analysts stressed, however, that these were only “in-sample” tests based on historical simulations, not actual trading results.