ETFs & Funds
Home›ETFs & Funds›Hedge Funds›Bracket22 builds an AI-only model to cut hedge fund la…
Bracket22 builds an AI-only model to cut hedge fund labor costs
The firm estimates its annual labor-related expenses fell from about $5m to as little as $40,000 after moving trading and research tasks to specialized AI agents.
Hedgeweek reports that hedge fund manager Brian Kelly has built Bracket22 around artificial intelligence agents instead of a traditional investment team, aiming to dramatically reduce labor-related expenses. According to CNBC, Kelly previously estimated annual labor costs at about $5m once salaries, healthcare, technology, office costs, and bonuses were included.
Bracket22, launched after Kelly experimented with AI in 2025, trades cryptocurrencies, equities, and commodities while investing only Kelly’s own capital. The business relies on a network of specialized AI agents to handle functions traditionally performed by analysts, traders, and quantitative researchers.
Kelly estimates that instead of $5m, the firm’s current AI infrastructure costs run between $30,000 and $40,000 a year, covering computing resources and the agents used to operate the strategy. He says the system is organized as specialists, with separate agents assigned to technical analysis and quantitative strategies.
Rather than one agent making every decision, Kelly designed the agents as specialists whose outputs are used to challenge his own thinking. He also estimates working with the AI agents has made him at least 10 times more productive, and he argues agentic AI could expand capacity for investment teams rather than only replacing staff.