JPMorgan-Built AI Agent Beats 60/40 Portfolio in Backtest

CoinFeed July 10 news, according to Bloomberg, JPMorgan researchers built multiple AI investment agents that can dynamically adjust allocations between stocks and bonds based on market conditions. Backtests show that over the past two decades, the best-performing AI agent had an annualized return 0.7 percentage points higher than a traditional 60/40 portfolio, with lower volatility, and also outperformed JPMorgan's own rule-driven market state model. The researchers noted that the results are from historical simulations, not actual investments, warning against viewing it as evidence that AI can consistently beat the market, and emphasized that 'agent AI needs to be based on a thoughtful asset allocation process, rather than naively assuming the agent itself can be a source of domain knowledge.'