Hedge fund manager Brian Kelly has built a trading firm around artificial intelligence agents rather than a traditional investment team, cutting annual labour-related costs from an estimated $5m to as little as $40,000, according to a report by CNBC.
Kelly, who previously ran a cryptocurrency hedge fund, launched Bracket22 after experimenting with AI in 2025. The firm trades cryptocurrencies, equities and commodities and invests only Kelly’s own capital.
The business operates around a network of specialised AI agents that perform functions traditionally handled by analysts, traders and quantitative researchers.
Kelly previously employed seven or eight people across different locations, including New York. Once salaries, healthcare, technology, office costs and bonuses were included, he estimates the firm’s annual labour-related expenses had reached about $5m.
By contrast, Kelly says his current AI infrastructure costs between $30,000 and $40,000 a year, including computing resources and the various agents used to operate the business.
The model illustrates the potentially dramatic impact agentic AI could have on the cost base of smaller hedge funds, while also raising questions about how much of a traditional investment operation can realistically be automated.
Kelly has assigned individual agents specific responsibilities. “Steffi” focuses on technical analysis, “Desmond” works on quantitative strategies, while “Houston” acts as a central control system coordinating the different functions.
Rather than allowing one system to make every decision, Kelly said he designed the agents as specialists and uses their outputs to challenge his own thinking.
Human judgement remains the final decision-making layer.
Kelly estimates that working with the AI agents has made him at least 10 times more productive. He argues, however, that the technology’s greatest potential may not be in eliminating employees altogether, but in significantly increasing the capacity of existing investment teams.
