OpenAI's Mercury Project Targets Wall Street Automation
OpenAI's Covert Mercury Project Aims to Automate Financial Modeling
OpenAI has quietly assembled a team of over 100 former bankers and financial experts for its internal "Mercury" project, according to Bloomberg sources. The initiative focuses on automating repetitive tasks traditionally performed by junior investment bankers, potentially disrupting core Wall Street operations.
Elite Team Composition
The Mercury project team includes:
- Former employees from Goldman Sachs, JPMorgan Chase, Morgan Stanley
- Experts from Brookfield, Evercore, and KKR
- MBA students from Harvard University and MIT
These specialists were recruited through third-party suppliers under flexible contracts paying approximately $150 per hour. Their primary responsibility involves creating simulated financial models for mergers, restructurings, or IPOs each week.
Innovative Training Methodology
The project employs a unique workflow:
- Experts write simple prompts to generate financial models
- AI outputs are converted into Microsoft Excel formats
- Models undergo continuous refinement based on feedback
- All operations follow strict industry standard protocols
The recruitment process itself demonstrates AI integration:
- Initial screening conducted by AI chatbot interviews (20 minutes)
- Followed by knowledge and modeling tests
- Submitted models enter OpenAI's training dataset directly
The company confirmed collaborating with domain experts through external suppliers to improve model capabilities.
Industry Implications
The Mercury project signals:
- AI's rapid advancement into specialized financial services
- Potential disruption of entry-level banking positions
- Shift toward automated financial modeling workflows
- Increased efficiency in deal analysis processes
The initiative represents one of the most concrete applications of conversational AI in high-stakes financial environments to date.
Key Points:
- OpenAI secretly developing Wall Street-focused AI automation
- Over 100 finance experts hired via third-party contracts
- Models trained using real-world merger/IPO simulations
- Recruitment and training heavily automated
- Potential to transform junior analyst roles
