OpenAI, the company behind ChatGPT, is quietly building what could be the biggest shake-up in modern finance jobs. Through a secret initiative called Project Mercury, OpenAI has reportedly hired more than 100 former investment bankers and consultants from top firms including Morgan Stanley, JPMorgan Chase, and Goldman Sachs — not to give them new banking jobs, but to teach artificial intelligence how to do theirs.
According to a Wall Street Journal report, the recruits are being paid about $150 an hour to train AI systems that can perform complex financial modelling — the kind of painstaking work usually handled by entry-level banking analysts.
The goal of Project Mercury is ambitious: to automate key analyst tasks such as building IPO models, restructuring analyses, and leveraged buyout projections. These time-consuming assignments often define the early years of a banking career, but OpenAI’s models are being designed to complete them in minutes — and with fewer human errors.
Former bankers involved in the project are helping refine these models by writing prompts, checking accuracy, and providing expert feedback. Consultants are also shaping the system’s ability to produce pitch decks and client-ready presentations, further encroaching on what has traditionally been the domain of fresh graduates entering finance.
While OpenAI describes the project as a way to enhance productivity, industry insiders see another story — one where entry-level banking roles could gradually disappear. If Project Mercury succeeds, much of the grunt work that trains analysts could be automated, leaving fewer opportunities for young professionals to break into investment banking.
The company’s broader enterprise strategy also points toward direct competition with established consulting and advisory firms. By combining engineers and domain experts, OpenAI is positioning itself as a new kind of AI-driven consultancy that can handle both data and decision-making
Morgan Stanley, already a key OpenAI partner in its Wealth Management division, is among the first to test Mercury’s prototypes. These trials merge the firm’s proprietary financial data with OpenAI’s generative models to create investment summaries, portfolio analyses, and client insights — all tasks that analysts once prepared manually.
Ironically, JPMorgan Chase — which recently extended a $4 billion credit line to OpenAI — is pursuing a similar vision of its own, aiming to become the world’s first “AI-powered megabank.”
Experts suggest that OpenAI’s latest move could mark a turning point for white-collar work. By embedding seasoned professionals into AI training, the company is shaping systems that understand the nuances of compliance, accuracy, and presentation — qualities that generic AI tools often lack.
If successful, Project Mercury could redefine the value of human labour in finance. The late-night grind of spreadsheets, models, and pitch decks — once a rite of passage for every young banker — might soon be handled by algorithms trained by those who lived that life first-hand.
