Hedge fund manager Brian Kelly keeps his staff working 24/7, but his labor costs are a fraction of what they used to be.
That’s because Kelly, who previously ran a cryptocurrency hedge fund, has founded a new trading company called Bracket22, which is fully powered by agent-based artificial intelligence.
“I used to have seven or eight employees all over the world, many of them based in New York,” Kelly told CNBC. “If you subtract their salaries, computing, medical expenses, etc., my salary is in the millions of dollars a year.”
Factoring in things like office space and bonuses, Kelly estimated that his total payroll costs before he started using AI were about $5 million a year.
“Right now, when I’m using AI, I’m making around ($30,000) to $40,000 a year in total, and that’s using all the AI agents, all my computing, everything I need to completely replicate a hedge fund…using AI,” he said.
Bracket22 is a stark example of the growing reality on Wall Street as companies test the benefits and limits of AI in finance.
JPMorgan Chase CEO Jamie Dimon said in February that technology was already reshaping the workforce and that his bank had plans for a “massive realignment” of its workforce. The company plans to launch an AI agent later this year, which it says can operate autonomously for hours at a time.
Morgan Stanley is similarly putting some of its work into AI.
However, there was some hesitation. For example, Goldman Sachs partners recently warned about the dangers of AI eroding bankers’ reasoning skills.
Kelly, a former trader on CNBC’s “Fast Money,” closed his crypto hedge fund in early 2025. Later that year, he began testing the use of artificial intelligence. Bracket22 invests only Kelly’s own capital and trades cryptocurrencies, stocks, and commodities.
Brian Kelly founded Bracket22, a trading company powered entirely by agent AI.
CNBC
Kelly introduced CNBC to several AI agents. Each agent had a unique role. Technical analysis is handled by a bot called “Steffi”. “Desmond” handles quantitative strategy, and “Houston”, fittingly, handles mission management and brings all the pieces together.
“I trained these agents to become specialists in their fields,” Kelly said. “I wanted to isolate them and give them an unbiased opinion on what I was doing.”
“And I will use my human judgment and human insight to make the final decision,” he added.
Kelly said he estimates productivity with agents is “at least 10 times better.” And while AI has replaced workers, the real opportunity lies in augmenting the human workforce, he said.
“If you hire 100 people, (with AI) you’ll have 1,000 people,” he said. “It doesn’t necessarily mean you can replace everyone with an AI agent. You can make your existing workforce at least 10 times more productive, and probably more.”
