OpenAI Hires Over 100 Former Bankers to Train AI for Financial Modeling

OpenAI has quietly launched an ambitious internal project codenamed "Mercury," aimed at automating the repetitive, foundational tasks typically performed by junior investment bankers on Wall Street. According to Bloomberg, the company has secretly hired more than 100 former bankers and financial experts to train its artificial intelligence systems in complex financial modeling. The project’s ultimate goal is to teach AI to independently generate sophisticated financial models—such as those used in mergers and acquisitions, restructurings, and initial public offerings—potentially eliminating hours of manual work traditionally done by entry-level analysts. The Mercury team is composed of elite professionals, including former employees from major financial institutions such as Goldman Sachs, JPMorgan Chase, Morgan Stanley, Brookfield, Evercore, and KKR, as well as MBA graduates from Harvard and MIT. These experts were recruited through third-party vendors and work on a flexible, project-based basis, earning approximately $150 per hour. Their primary responsibility involves building a simulated financial model each week—based on real-world transaction types—using standard industry practices. They then use simple prompts to guide the AI, convert the model output into Microsoft Excel, and refine it based on feedback. All work is done in strict compliance with professional financial modeling standards. The recruitment and training process for the project is largely automated. Candidates first undergo a 20-minute interview conducted by an AI chatbot, followed by assessments in financial knowledge and modeling. The models they produce are reviewed, and the feedback is directly fed into OpenAI’s training data. This iterative process is designed to help the AI system learn to generate accurate, professional-grade financial models on its own. An OpenAI spokesperson confirmed the company collaborates with experts across various fields to improve and evaluate its models, noting that these professionals are hired and compensated through external vendors. The spokesperson emphasized that the work is part of ongoing efforts to enhance the AI’s real-world applicability. The emergence of the Mercury project signals a significant shift: conversational AI is no longer limited to general-purpose tasks. It is rapidly advancing into highly specialized domains like finance, where precision and expertise are paramount. As a result, the project raises serious questions about the future of entry-level investment banking roles, as AI systems become capable of handling the foundational work that has long defined the job.