Key Takeaways
- • Brian Kelly replaced a $5 million annual payroll with AI agents costing $30,000–$40,000 per year — a 99% reduction in labor costs while maintaining 24/7 trading operations across cryptocurrencies, stocks, and commodities through his firm Bracket22.
- • Kelly estimates 10x productivity gain using agentic AI — a significant efficiency multiplier that reflects the autonomous capability of AI agents to operate continuously without geographic or time constraints, compared to his previous global trading staff of 7–8 employees.
- • Bracket22 operates entirely on proprietary capital with no external investors, allowing Kelly to run an experimental, AI-first hedge fund model that would otherwise face institutional guardrails or compliance scrutiny typical of multi-investor funds.
- • Major Wall Street firms are racing to deploy similar AI infrastructure — JPMorgan Chase and Morgan Stanley are actively integrating autonomous AI agents into trading and operations, signaling industry-wide adoption beyond Bracket22's proof-of-concept model.
- • Cost structure demonstrates AI's structural impact on finance employment — reducing overhead from office space, salaries, healthcare, bonuses, and global coordination to compute and licensing fees represents a fundamental shift in operational economics for trading firms.
Summary
Brian Kelly, a former CNBC "Fast Money" trader, has constructed Bracket22 as an operational blueprint for AI-powered hedge fund management. After closing his cryptocurrency hedge fund in early 2025, Kelly began architecting a trading operation entirely staffed by agentic artificial intelligence rather than human traders and analysts. The firm invests exclusively Kelly's personal capital and executes trades across cryptocurrencies, stocks, and commodities through autonomous AI systems operating continuously.
The economic transformation is stark. Kelly's previous trading operation maintained 7–8 employees distributed globally, primarily in New York, with total annual labor-related costs—including salaries, healthcare, bonuses, compute resources, and office infrastructure—reaching approximately $5 million annually. Under the new AI-agent model, Bracket22's total operational cost, encompassing all AI agents, compute infrastructure, and necessary technology stack, runs between $30,000 and $40,000 per year. This represents a reduction of over 99% in staffing costs while enabling 24/7 market participation without geographic or shift constraints. Kelly characterizes the productivity gain at "at least 10 times" more efficient than his previous human-staffed operation.
The Bracket22 model reflects a broader Wall Street trend of deploying autonomous AI capabilities. JPMorgan Chase CEO Jamie Dimon stated in February 2026 that AI was already reshaping the bank's workforce, with planned "huge redeployment" of employees and autonomous AI agents capable of working for hours without human intervention scheduled for later in the year. Morgan Stanley has similarly begun routing operational work toward AI systems. However, skepticism persists within finance leadership; a Goldman Sachs partner recently cautioned against allowing AI dependency to erode bankers' foundational reasoning skills.
Bracket22's structure—operating solely on Kelly's capital rather than institutional investor money—provides latitude for this experimental deployment that would face regulatory and governance scrutiny in multi-investor fund structures. The firm's success or failure will serve as a critical data point for the financial industry as it calculates the true ROI of replacing human trading infrastructure with agentic AI systems.
The economic transformation is stark. Kelly's previous trading operation maintained 7–8 employees distributed globally, primarily in New York, with total annual labor-related costs—including salaries, healthcare, bonuses, compute resources, and office infrastructure—reaching approximately $5 million annually. Under the new AI-agent model, Bracket22's total operational cost, encompassing all AI agents, compute infrastructure, and necessary technology stack, runs between $30,000 and $40,000 per year. This represents a reduction of over 99% in staffing costs while enabling 24/7 market participation without geographic or shift constraints. Kelly characterizes the productivity gain at "at least 10 times" more efficient than his previous human-staffed operation.
The Bracket22 model reflects a broader Wall Street trend of deploying autonomous AI capabilities. JPMorgan Chase CEO Jamie Dimon stated in February 2026 that AI was already reshaping the bank's workforce, with planned "huge redeployment" of employees and autonomous AI agents capable of working for hours without human intervention scheduled for later in the year. Morgan Stanley has similarly begun routing operational work toward AI systems. However, skepticism persists within finance leadership; a Goldman Sachs partner recently cautioned against allowing AI dependency to erode bankers' foundational reasoning skills.
Bracket22's structure—operating solely on Kelly's capital rather than institutional investor money—provides latitude for this experimental deployment that would face regulatory and governance scrutiny in multi-investor fund structures. The firm's success or failure will serve as a critical data point for the financial industry as it calculates the true ROI of replacing human trading infrastructure with agentic AI systems.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.