OpenAI enlists more than 100 former investment bankers to help train its artificial intelligence in building financial models, aiming to replace the hours of manual labor performed by junior analysts across the industry.
The group, which includes former employees of JPMorgan Chase & Co., Morgan Stanley and Goldman Sachs Group Inc., is part of a confidential project within the startup, codenamed Mercury, according to documents obtained by Bloomberg.
Participants are paid $150 an hour to write prompts and build financial models for various types of transactions, including restructurings and initial public offerings, according to a person familiar with the project. The company also gave hires early access to the AI it is developing, which aims to replace entry-level tasks at investment banks.
Take your business to the next level with the country’s top entrepreneurs!
The project highlights the urgency for OpenAI, led by Sam Altman, to make its powerful AI technology more useful for companies across sectors such as consulting, finance, legal and technology. Despite reaching a $500 billion valuation earlier this month, the world’s biggest startup has yet to turn a profit.
An OpenAI spokesperson said the company works with a variety of experts “to improve and evaluate the capabilities of our models across different domains. Experts are recruited, managed, and compensated by third-party vendors.”
Investment banking analysts typically spend more than 80 hours a week at their desks, working on live deals and building detailed models in Microsoft’s Excel for mergers and leveraged acquisitions. They often receive a series of requests for tweaks to PowerPoint presentations, creating a culture that gave rise to the “pls fix” meme on Wall Street.
There are already several startups looking to equip banks with AI to help with these tasks. Although analysts complain about the exhausting routine, the advancement of AI also raises concerns about the security of their jobs.
The selection process for Project Mercury involves almost no human interaction, according to the source who requested anonymity to discuss non-public information.
The first step is an interview lasting around 20 minutes with an AI chatbot, which asks questions based on the candidate’s resume. The second phase tests knowledge about financial statements. The final step is a modeling test.
Continues after advertising
The work is flexible and contractors must deliver one model per week, according to the source. Instructions include writing prompts in simple terms and then running the model. Participants receive feedback from a reviewer and must correct any problems before their work is integrated into OpenAI systems.
So far, Project Mercury has attracted participants who previously worked at several Wall Street financial centers, including Brookfield Corp., Mubadala Investment Co., Evercore Inc. and KKR & Co., the documents show. Some MBA candidates at Harvard and MIT also participate in the project.
Participants must create their templates in Excel and follow industry standards for formatting, including margins and italics for percentages.
Continues after advertising
© 2025 Bloomberg L.P.
