According to reports from Business Standard and Bloomberg, JPMorgan Chase & Co researchers built a range of AI-powered investing agents that shift between stocks and bonds depending on changing market conditions. In backtests covering the past two decades, the best-performing system outperformed a traditional 60/40 portfolio by 0.7 percentage point annually while delivering 2.8% lower annual volatility. The system also outperformed JPMorgan's own rules-based market regime model, as reported by strategists led by Thomas Salopek. As per NDTV Profit, the results come with an important caveat - they are based on historical simulations rather than live investing, and JPMorgan warns against treating them as proof that AI can consistently outperform markets. The findings were announced in a Thursday, 9 July note by JPMorgan strategists, offering an early look at how Wall Street is testing artificial intelligence for one of finance's hardest calls: where to put money.
As reported by Business Standard and Bloomberg, the JPMorgan team designed a system using agents powered by models from OpenAI and Anthropic. The system classifies markets into four regimes based on growth and inflation: Goldilocks, reflation, stagflation and risk-off. The AI agents were then tasked with deciding how to allocate capital across asset classes in each environment, favouring equities during periods of strong growth and increasing fixed-income exposure as outlook deteriorated. According to NDTV Profit, this represents JPMorgan's first attempt to build an AI system for identifying market regimes, with the strategists describing the work as "the firm's first attempt to build an AI system for identifying market regimes." The agents were designed to identify whether the market backdrop favoured risk-taking or caution, with the system able to favour equities in strong growth environments and move toward bonds when outlook weakened or inflation pressures changed.
According to the Business Standard and Bloomberg reports, all eight AI agents tested outperformed the traditional 60/40 portfolio on a risk-adjusted basis. The results demonstrate how AI technology can improve upon existing frameworks, as the agents were able to beat JPMorgan's existing rules-based market regime model. This suggests the technology was able to enhance a framework already used to guide asset-allocation decisions. As per NDTV Profit, the findings add to a growing body of evidence suggesting AI can perform increasingly sophisticated investment tasks, with the JPMorgan strategists noting that "the AI agent can be set up with a process to be empowered to make decisions under uncertainty, producing outperformance vs a reasonable benchmark." The attractive part of the test lies in acknowledging that financial markets are full of strategies that looked brilliant in simulation but failed once exposed to real capital, transaction costs, crowding and changing investor behavior.
As reported by NDTV Profit, the experiment offers an early glimpse of Wall Street's next phase of AI adoption, with banks spending the past two years embedding large language models into research, coding and internal investment tools. The findings add to growing evidence that AI can perform increasingly sophisticated investment tasks, though JPMorgan cautioned against treating the results as proof that AI can consistently outperform markets. The JPMorgan strategists also acknowledged broader risks, writing that "We strongly caution against uncritically accepting what amounts to in-sample, overly confident answers of AI." They emphasized that "Agentic AI needs to be grounded in a well thought-out asset allocation process, rather than naively assuming the agent can be the source of the domain knowledge." The study matters because asset allocation sits at the center of investment management, but the hurdle remains high as JPMorgan's own warning suggests "agentic AI may help structure decisions, but it still needs human oversight, a disciplined investment process, and live-market proof before it can be trusted with capital allocation at scale."
