Why OpenAI is hiring 100 ex-bankers: Inside the ChatGPT-maker's secret project to automate Wall Street's grunt work
OpenAI has enlisted over 100 former Wall Street investment bankers to train its artificial intelligence systems in building complex financial models, as part of a secret project called Mercury.
ADVERTISEMENT
OpenAI is turning to former Wall Street talent to teach its artificial intelligence how to perform some of investment banking’s most time-consuming tasks.
According to a Bloomberg report, the Sam Altman–led company has hired more than 100 ex-bankers from firms such as JPMorgan Chase & Co., Morgan Stanley, and Goldman Sachs Group Inc. to train its AI models to build financial models for transactions including IPOs, restructurings, and leveraged buyouts.
The project, internally codenamed Project Mercury, aims to automate the grunt work typically handled by entry-level analysts at major investment banks. Participants in the program are reportedly paid $150 an hour to write prompts, test models, and provide feedback on AI-generated outputs.
Bloomberg reported that contractors also get early access to OpenAI’s financial modeling tools, which are being designed to replicate, and eventually streamline, the painstaking manual work done by junior bankers.
Despite achieving a $500 billion valuation earlier this month, OpenAI has yet to turn a profit. The company is pushing to make its technology more commercially useful across industries including finance, consulting, law, and tech.
Investment banking analysts typically spend long hours building detailed Excel models and editing presentation decks for senior executives. The repetitive nature of the work has long been a point of frustration, reflected in Wall Street’s “pls fix” meme culture. Now, AI’s entry into this space is sparking both curiosity and concern over job security among junior bankers.
The application process for Project Mercury is also largely AI-driven. Applicants undergo a 20-minute interview with a chatbot, followed by financial statement and modeling tests. Contractors are required to submit one model per week, receiving feedback before their work is integrated into OpenAI’s systems.
