OpenAI is running an ambitious internal project aimed at automating the repetitive tasks junior investment bankers typically handle. According to a Bloomberg report, the company has hired more than 100 former bankers who are using their financial modeling experience to train AI systems. The effort operates under the codename "Mercury."
The team includes former employees from major firms like Goldman Sachs, JPMorgan, Morgan Stanley, Brookfield, Evercore, and KKR, along with MBA students from Harvard and MIT. They’re recruited through third-party vendors, work on flexible schedules, and earn about $150 an hour. Each week, participants build a financial model that simulates transactions such as mergers, restructurings, or IPOs. They write simple prompts, translate the results into Microsoft Excel, and refine the models based on feedback - all following standard industry formatting.
Recruitment for the project is almost entirely automated. As Bloomberg’s source explained, candidates start with a 20-minute interview conducted by an AI chatbot, then complete knowledge and modeling tests. Their finished models are reviewed, and the reviewers’ feedback goes directly into OpenAI’s training data. The knowledge gained from these examples is meant to teach the AI how to produce financial models on its own, potentially eliminating some of the tedious groundwork junior analysts spend hours on.
An OpenAI spokesperson told Bloomberg the company works with experts "across various domains to improve and evaluate our models’ capabilities," noting that these experts are recruited and paid by outside vendors.
