Hedge-fund manager Brian Kelly has fundamentally transformed his trading operations by building Bracket22, a firm powered entirely by artificial intelligence agents. After previously managing a cryptocurrency hedge fund with a staff of seven or eight employees and annual labor costs reaching about $5 million, Kelly now operates his firm with AI agents at a fraction of the cost, spending around $30,000 to $40,000 annually. These AI agents work around the clock, efficiently replicating the functions of a traditional hedge fund without the overhead of human staff.
Bracket22’s AI agents are specialized for different roles within the firm. One, dubbed “Steffi,” focuses on technical analysis, while another, “Desmond,” manages quantitative strategies. “Houston” serves as a central coordinator, pulling all the pieces together. Kelly emphasizes that he uses these AI agents to obtain unbiased insights and then applies his human judgment to make final trading decisions, blending cutting-edge technology with personal expertise.
The transition at Bracket22 highlights a broader trend on Wall Street where major financial institutions like JPMorgan Chase and Morgan Stanley are increasingly integrating AI agents into their workflows. JPMorgan plans to deploy AI agents capable of working autonomously for hours, signaling significant workforce redeployment. However, the adoption of AI is not without caution; some industry leaders, including a Goldman Sachs partner, have expressed concern about potential impacts on bankers’ critical reasoning skills.
Kelly estimates his productivity has increased at least tenfold due to utilizing AI agents. Rather than replacing humans outright, he sees the technology as a powerful tool to dramatically amplify employee productivity. Kelly envisions a future where AI can expand a workforce’s capabilities, multiplying output as if turning a team of 100 into a team of 1,000, thus enhancing efficiency without completely eliminating human involvement in the trading process.
Start the discussion with a take, question, or market read.