Former cryptocurrency hedge fund manager Brian Kelly has founded Bracket22, a trading firm powered entirely by agentic artificial intelligence, slashing annual labor-related operating costs from approximately $5 million down to between $30,000 and $40,000 as Wall Street increasingly tests autonomous financial agents.
The $5 Million Labor Model Replaced by Code
When financial markets open, former trader on CNBC’s “Fast Money” Brian Kelly operates an entirely autonomous hedge fund from his desk. After shuttering his previous cryptocurrency hedge fund in early 2025, Kelly launched Bracket22 later that year to trade stocks, commodities, and cryptocurrencies using exclusively his own capital. According to reports from CNBC, the operational blueprint of the firm relies on specialized software rather than human capital.
Here is the math. Kelly previously maintained a global footprint of seven to eight employees primarily stationed in New York. Between base salaries, health insurance, computing resources, bonuses, and office space, his total overhead reached roughly $5 million annually. Today, running Bracket22 with agentic AI cuts that expenditure significantly. Total annual costs—covering every specialized AI agent, compute, and everything needed to replicate a hedge fund—run between $30,000 and $40,000.
The Bottom Line
- Dramatic Cost Compression: Brian Kelly reduced Bracket22’s annual operational overhead from approximately $5 million to a maximum of $40,000 by replacing a 7-to-8 person human staff with autonomous AI agents.
- Specialized Agent Architecture: The firm operates via distinct software units—”Steffi” for technical analysis, “Desmond” for quantitative strategies, and “Houston” as mission control—while retaining human judgment for final execution.
- Broader Wall Street Adoption: Major institutions are rapidly following suit; JPMorgan Chase (NYSE: JPM) CEO Jamie Dimon announced workforce restructuring plans and autonomous agent rollouts earlier this year, aligning with similar tests at Morgan Stanley (NYSE: MS).
Inside Bracket22’s Autonomous Agent Hierarchy
Rather than relying on a single large language model to manage portfolio risk, Kelly engineered a modular division of labor among his artificial intelligence systems. Each agent fulfills a rigid analytical mandate designed to eliminate emotional bias from the trading process.
The system relies on three core entities introduced by Kelly in public demonstrations. An agent named “Steffi” monitors charting formations and executes technical analysis. Another agent, “Desmond,” runs quantitative strategies. Finally, an agent designated as “Houston” acts as mission control, synthesizing data streams across asset classes.
“I’ve crafted each of these agents to be a specialist in their field,” Kelly noted in interviews with CNBC. “I wanted to isolate them and I wanted to get their unbiased view on what I’m doing.” Once the artificial intelligence components aggregate market intelligence, Kelly applies his own discretionary oversight to make the ultimate capital allocation decision.
| Operational Metric | Traditional Hedge Fund (Pre-2025) | Bracket22 AI Model |
|---|---|---|
| Headcount | 7 to 8 Global Employees | 0 Full-Time Staff |
| Annual Overhead Costs | ~ $5,000,000 | $30,000 to $40,000 |
| Capital Deployed | Not Specified | Proprietary Capital Only |
| Asset Classes | Cryptocurrencies | Cryptocurrencies, Equities, Commodities |
Wall Street’s Great Workforce Redirection
Bracket22’s lean structure highlights a broader operational pivot across traditional global banking.
At JPMorgan Chase (NYSE: JPM), executives are preparing to launch autonomous AI agents capable of executing multi-hour analytical workflows independently. CEO Jamie Dimon noted during a February address that the technology is actively reshaping the bank’s internal structure, accompanied by comprehensive workforce redeployment plans. Similarly, Morgan Stanley (NYSE: MS) continues to channel operational tasks into machine-learning frameworks to accelerate client services.
However, pushback persists within elite financial circles. A partner at Goldman Sachs (NYSE: GS) recently cautioned against the risk of artificial intelligence eroding essential critical reasoning skills among bankers.
Scaling Productivity Without Expanding Headcount
Despite stripping his payroll down to near zero, Kelly argues that the ultimate promise of financial artificial intelligence is not mass corporate displacement, but extreme individual leverage. He estimates his personal operational output has expanded by a factor of ten since deploying his custom agent architecture.
“If you take a staff of 100, [with AI] you’ve got a staff of a thousand,” Kelly explained. “It’s not necessarily just, hey, you can replace everybody with AI agents. You can make your existing employees at least 10 times — maybe more — more productive.” As quantitative funds and proprietary trading desks evaluate performance metrics, Bracket22 stands as a working proof-of-concept for ultra-low-cost, software-driven asset management.
Disclaimer: The information provided in this article is for educational and informational purposes only and does not constitute financial advice.
