JPMorgan tests AI agents for portfolio management: outperformed 60/40

JPMorgan Chase & Co. is testing AI agents capable of independently managing an investor's portfolio. In simulations using historical data, these agents demonstrated results exceeding the classic 60/40 portfolio — a strategy that for many decades has been considered the gold standard of conservative investing.

How Investment AI Agents Work

The system uses machine learning to analyze market data and make decisions about asset allocation across different types of securities. Unlike traditional investment algorithms, JPMorgan's AI agents can adapt to changing market conditions in real time and rebalance the portfolio without human portfolio manager involvement.

Investors have been using AI for auxiliary tasks for several years: company analysis, risk assessment, and market anomaly detection. JPMorgan is going further and delegating a key function to its neural network — making independent decisions about how to distribute funds between stocks, bonds, and other assets.

This represents a shift in AI's role from assistant to autonomous investor, though for now only in testing conditions.

The 60/40 Portfolio as the Primary Benchmark

JPMorgan is testing this technology using backtests — simulations on historical data. The results showed that AI agents outperform the traditional 60/40 portfolio (60% stocks, 40% bonds).

The 60/40 portfolio is a legendary investment strategy recommended to conservative investors, beginning traders, and even wealthy individuals seeking to preserve capital with minimal risk. It provides a balance between growth (stocks typically grow over the long term) and protection (bonds are more stable). This approach has demonstrated reliable results for decades.

• AI agents outperformed the 60/40 portfolio in backtests
• The system can independently rebalance assets in real time
• JPMorgan is developing the technology for implementation in client portfolio management
• This demonstrates a trend: autonomous systems are assuming functions previously performed by portfolio managers

An

Important Caveat: Historical Data Does Not Guarantee the Future

Here is a significant caveat: success in backtests never guarantees success in live markets. Markets constantly change, and new geopolitical factors and shocks emerge that did not occur in the past. Many strategies that work perfectly on historical data fail when confronted with reality.

Machine learning can overfit to historical patterns that will never repeat. JPMorgan surely accounts for this problem and is testing AI precisely to evaluate the reliability of results before launching for real client money.

What This Means for the Financial Industry

JPMorgan is one of the world's largest investment banks. Its decision to experiment with AI for autonomous portfolio management shows that neural networks are gradually assuming functions that were previously performed by human portfolio managers. If JPMorgan ultimately launches this system for real clients, it could become a significant milestone in the history of the financial industry.