OpenAI, the artificial intelligence company behind ChatGPT, is quietly training its systems to perform the kind of routine but essential tasks that occupy most junior bankers’ time, Bloomberg reported.
Discover how neo-banks become wealthtech in London at the fmls25
Inside the Secretive “Project Mercury”
Documents cited by the media publication reveal that OpenAI has enlisted more than 100 former investment bankers for an internal project called Mercury.
The initiative brings together ex-employees from JPMorgan Chase, Morgan Stanley, Goldman Sachs, and other financial giants. Their mission: to teach AI how to build financial models for restructurings, IPOs, mergers, and leveraged buyouts.
- OpenAI and Broadcom Ink 10-Gigawatt Chip Deal to Feed AI Hunger
- OpenAI’s New Structure Aims to Attract Billions While Preserving Mission
- Is OpenAI the Next Big Social Media Player? Sam Altman Teases Bold Expansion
Participants in the project are paid about $150 an hour to write prompts, build models, and refine the AI’s performance. A person familiar with the program said these contractors also get early access to OpenAI’s developing technology that aims to replicate entry-level work in finance.
A spokesperson for OpenAI confirmed to Bloomberg that the company collaborates with “a range of experts to improve and evaluate the capability of our models across different domains,” adding that these specialists are “recruited, managed, and compensated by third-party suppliers.”
The initiative highlights OpenAI’s drive to turn its AI prowess into practical tools for industries that rely heavily on repetitive data work. For investment banking, that could mean automating the kind of Excel modeling and presentation building that keeps analysts working upwards of 80 hours per week.
Aiming to Cut the Grind, Not the Quality
The culture of constant revisions – epitomized by Wall Street’s “pls fix” meme – has long defined the analyst experience. Now, AI could soon take over those endless adjustments, sparking equal parts optimism and anxiety within the industry.
Getting involved in the project reportedly requires passing through a hiring process run almost entirely by AI.
Applicants complete a 20-minute chatbot interview, a financial statement test, and a modeling exercise. Contractors are expected to deliver one financial model per week, using standard industry formats and Excel conventions such as margin spacing and italicized percentages.
Read more: Fed's Governor Proposes Access to Central Bank’s Payment Infrastructure for Crypto Firms
After each submission, a reviewer provides feedback before the model is integrated into OpenAI’s training systems. The setup has attracted professionals from firms including Brookfield, Mubadala, Evercore, and KKR, as well as MBA students from Harvard and MIT.
OpenAI’s $500 billion valuation underscores its dominance in the AI space, but profitability remains elusive. Project Mercury signals a broader strategy: transforming industries like finance, law, and consulting through automation that goes beyond chat interfaces.
ChatGPT 5 and the Trading and Investment Space
A few months ago, OpenAI launched GPT-5, its most advanced artificial intelligence model yet, marking a major milestone in AI development. The new model is available to all users – including those on the free tier – broadening access to cutting-edge AI capabilities.
Microsoft CEO Satya Nadella praised the release on X, describing it as “the most capable model yet from our partners at OpenAI.” AI researcher Dave Wang also highlighted GPT-5’s potential to transform finance and investing, pointing to significant improvements in speed, accuracy, and reasoning.
The surprise rollout underscored OpenAI’s ambition to make generative AI a mainstream tool across industries. The company said GPT-5 has been optimized for a wide range of applications, from writing and coding to education and healthcare.
