JPMorgan Chase & Co. is pushing beyond AI-assisted stock picking into full portfolio allocation, testing models that decide how to distribute capital across asset classes. Early backtests show these agents beating the traditional 60/40 stock-bond split, a benchmark that has anchored institutional and retail portfolios for decades. The move signals a shift from using AI as a research tool to deploying it as an autonomous allocator, potentially upending the $130 trillion global asset management industry.

The 60/40 model has struggled recently as correlations between equities and fixed income rose during 2022's rate-hike cycle, delivering its worst annual return since the 1930s. JPMorgan's AI approach aims to dynamically adjust exposures based on regime detection, liquidity conditions, and cross-asset signals — tasks that currently require teams of strategists and risk managers. If validated out-of-sample, the technology could compress fees, reduce human bias, and accelerate the shift toward systematic, data-driven allocation.

Regulatory scrutiny looms. The SEC and Fed have flagged AI-driven trading and advisory models for opaque decision-making and systemic risk. JPMorgan must demonstrate explainability and robust guardrails before client capital follows backtest results. Competitors including BlackRock and Morgan Stanley are pursuing similar capabilities, setting up an arms race where the winner defines the next generation of portfolio construction.