Hedge Fund Manager Cuts Costs From $5M to $40K Using AI Agents

Former hedge fund manager slashes operating costs with AI workforce

Brian Kelly has eliminated nearly all his payroll expenses by building a trading firm that operates exclusively through artificial intelligence agents. The former cryptocurrency hedge fund manager now runs Bracket22 with annual costs of roughly $30,000 to $40,000—down from approximately $5 million when he employed seven to eight human staff members.

The dramatic cost reduction includes all AI subscriptions, computing infrastructure, and operational needs required to replicate traditional hedge fund functions. Kelly's previous expenses covered salaries, healthcare, office space, and bonuses for employees distributed globally, with many based in New York.

Bracket22 trades Kelly's personal capital across cryptocurrencies, stocks, and commodities using specialized AI agents. Each bot handles a distinct function: "Steffi" performs technical analysis, "Desmond" executes quantitative strategies, and "Houston" coordinates the overall operation. Kelly designed each agent as a specialist to provide isolated, unbiased perspectives on trading decisions.

Why it matters

Kelly's operation represents an extreme test case for AI's economic impact on financial services. While major banks like JPMorgan Chase and Morgan Stanley are deploying AI agents to augment human teams, Bracket22 demonstrates the technology's potential to eliminate traditional staffing models entirely. The cost differential—a 99% reduction—provides concrete data for firms evaluating AI investment returns, though questions remain about scalability, regulatory compliance, and risk management in fully automated trading environments.

Wall Street's broader AI adoption

Major financial institutions are moving cautiously toward similar automation. JPMorgan Chase CEO Jamie Dimon stated in February that AI was already reshaping the bank's workforce, with plans to launch autonomous AI agents later this year capable of working independently for extended periods. The bank has indicated significant employee redeployment initiatives are underway.

Morgan Stanley has likewise begun directing certain work functions to AI systems. However, adoption isn't universal—a Goldman Sachs partner recently raised concerns about AI potentially degrading bankers' analytical and reasoning capabilities.

Human judgment remains central

Despite eliminating his human staff, Kelly maintains that his role involves final decision-making authority. He reviews the specialized analysis from each AI agent and applies human judgment to trading choices. Kelly estimates he operates at least ten times more productively with his AI infrastructure than he did with traditional employees.

Kelly argues the technology's greatest value lies in workforce augmentation rather than replacement. He suggests organizations with 100 employees could effectively scale to 1,000-person productivity levels by deploying AI agents alongside human workers, potentially multiplying individual employee output by a factor of ten or more.

Kelly, previously a trader on CNBC's "Fast Money," closed his cryptocurrency hedge fund in early 2025 and began experimenting with AI applications shortly afterward before launching Bracket22.

These details were first reported by CNBC.