JPMorgan Chase & Co. is testing whether artificial intelligence “agents” can do more than analyze markets—specifically, whether they can allocate capital in response to changing conditions. According to reporting cited by Bloomberg and repeated by other outlets, the bank has been running historical simulations in which AI agents decide how to distribute investment between stocks rather than relying solely on fixed rules.
Across the backtests described, eight AI agents are reported to outperform both a traditional 60/40 portfolio benchmark and JPMorgan’s own rules-based market regime model. The results are attributed to the bank’s strategists, with Thomas Salopek noted as leading the work. The firms’ findings are presented as an early indication of how AI could be incorporated into asset allocation and risk management, as investor interest in AI tools grows.
The sources characterize the exercise as backtesting rather than live performance, and they do not provide details in the excerpts on time periods, risk constraints, or magnitude of the outperformance.
