San Francisco: OpenAI has reportedly onboarded more than 100 former investment bankers for a unique initiative called Project Mercury, which aims to train its artificial intelligence systems to handle routine financial tasks traditionally performed by junior bankers.
Former Wall Street professionals on board
The project, according to sources familiar with its development, involves former employees from some of the world’s top financial institutions — including JPMorgan Chase, Morgan Stanley, and Goldman Sachs. These finance professionals are paid about $150 per hour to create financial models and prompts that help OpenAI’s systems learn the intricate workflows of investment banking.
Their work covers complex tasks such as building models for mergers and acquisitions (M&A), IPOs, buyouts, and restructuring deals. The AI models are being trained not only to execute financial calculations but also to replicate presentation-ready formatting — from Excel layouts to PowerPoint adjustments that investment banks rely on.
Goal to automate junior banking tasks
The core goal of Project Mercury is to automate much of the time-consuming “grunt work” handled by junior analysts and associates, who often spend 80–100 hours per week managing financial models, adjusting slides, and responding to client edits.
This could represent a major shift in the finance industry, where automation has already started replacing entry-level roles in data analysis and market research. For OpenAI, the project reflects its push to make AI tools applicable beyond tech — targeting sectors such as finance, consulting, and law.
An OpenAI spokesperson said the company works with domain experts “to improve and evaluate the capability of our models across different fields.” These experts, the company clarified, are hired through third-party suppliers and do not represent direct OpenAI employees.
Streamlined AI-driven hiring
The hiring process for Project Mercury itself is partly automated. Applicants begin with a 20-minute AI chatbot interview, followed by assessments in financial modelling and statement analysis. Selected candidates are required to build one financial model per week, incorporating detailed feedback before integration into OpenAI’s training systems.
Participants include professionals from firms like Brookfield, Evercore, Mubadala Investment, and KKR, as well as MBA students from top business schools such as Harvard and MIT.
Each model must adhere to industry presentation standards, such as correct formatting, consistent font sizes, and margin adjustments — ensuring the AI learns not only the financial logic but also the stylistic expectations of Wall Street deliverables.
Implications for the future of banking
While Project Mercury may enhance productivity and accuracy in financial analysis, it has also stirred anxiety among young professionals in banking. As AI becomes increasingly proficient in financial modelling and report generation, concerns are growing that it could make junior banking roles obsolete.
Despite such worries, experts believe AI could also create new opportunities — shifting focus from manual modelling to strategic analysis and client management.
For OpenAI, whose valuation now exceeds $500 billion, Project Mercury signifies a step toward commercialising its technology for enterprise use and developing a sustainable, profitable business model