Countdown to the end of “technical worker” careers in finance? OpenAI employs over 100 former investment bank staff to train models, aiming to have AI replace “junior investment banking employees.”

```

Artificial intelligence company OpenAI is conducting an extraordinary experiment: training its AI models in core skills such as financial modeling by hiring more than 100 former investment banking employees.

According to media reports on the 22nd, OpenAI's secret project, codenamed "Mercury", has already attracted participation from former employees of top financial institutions such as JPMorgan Chase, Morgan Stanley, and Goldman Sachs. These participants are responsible for writing prompts and building financial models for various transactions such as restructuring and initial public offerings (IPOs), in order to "teach" AI how to work like a junior banker.

Citing informed sources, these contractors are paid an hourly wage of $150 and are given early access to the AI they are helping to develop. This move has sparked concerns within the financial industry about the career prospects of junior positions. Analysts have long complained about heavy and tedious workloads, and the rise of AI is now turning these complaints into a direct threat to their job security.

This move also reveals the strategic focus of OpenAI under Sam Altman's leadership. As one of the world's highest-valued startups, OpenAI has yet to turn a profit. Converting powerful AI technology into practical business tools that serve industries such as finance, consulting, and law has become an urgent priority.

More than 100 former employees participate, training AI at $150/hour

According to reports, OpenAI's "Mercury" project has recruited more than 100 professionals with Wall Street backgrounds. The participant lineup is impressive, including not only former employees of leading investment banks such as JPMorgan Chase, Morgan Stanley, and Goldman Sachs, but also talent from well-known investment institutions such as Brookfield Corp., Mubadala Investment Co., Evercore Inc., and KKR & Co., and even current MBA students from Harvard University and MIT.

The project operates in a flexible contracting model, requiring participants to submit a financial model each week. Their tasks include writing prompts in simple language and then executing and building models in Microsoft Excel. A reviewer provides feedback, and participants must correct all issues before their work is ultimately submitted to OpenAI's system. A spokesperson for OpenAI said the company "works with experts from many fields to improve and evaluate our models' capabilities," and that these experts are recruited, managed, and paid by third-party vendors.

It is worth noting that the project's application process is almost entirely automated. The process includes a roughly 20-minute interview with an AI chatbot, a financial statement knowledge test, and a final modeling skills assessment.

"Please fix": Details and formatting that AI needs to learn

Junior investment banking analysts are known for their extremely long working hours, often exceeding 80 hours a week during live deals. Their work is filled with what are considered "dirty and strenuous tasks," such as building complex M&A models in Excel and repeatedly revising PowerPoint presentations as requested by superiors.

In Bloomberg columnist Matt Levine's view, this extreme attention to detail is precisely what AI needs to learn. He commented that a successful investment banking analyst must be "detail-oriented," because even a minor formatting error in a model or presentation—such as dollar signs not aligning—can make a boss suspect more serious substantive mistakes are present, undermining trust.

Levine compares current generative AI to a "clever but careless analyst"—deft at quickly outputting seemingly reasonable models, but often prone to mistakes or "hallucinations," making them hard to fully trust. Therefore, he believes OpenAI's project is essentially a form of "reinforcement learning," equivalent to "having an investment banking VP yell at the AI for years until it gets the formatting right." Hiring former bankers targets precisely their own experience enduring this rigorous training, making them instinctively adhere to industry norms like margin size and italicizing percentages.

Replacing apprenticeships? Wall Street’s promotion ladder under test

The direct goal of the "Mercury" project is to let AI replace junior staff jobs, which raises profound questions about the future of the traditional apprenticeship model in investment banking. For decades, Wall Street has relied on a pyramid structure: junior analysts learn skills by handling basic chores, with the best eventually promoted to senior bankers who interact with clients.

As Matt Levine asks: When an industry no longer needs apprentices, where will its future senior partners come from? If AI takes over model building and paperwork, how will young finance professionals acquire the hands-on experience and industry knowledge necessary to become tomorrow’s leaders?

However, there is another side to this trend. Analyst programs in investment banking have always had high staff turnover, with many choosing to leave after two years to pursue entrepreneurship or other fields. For those who have already left, coming back to train a robot that can replace their former position may not carry much psychological burden.

The "Mercury" project is not only a potential disruptor for Wall Street, but also a microcosm of OpenAI's commercialization strategy. By targeting the lucrative field of financial services, OpenAI is striving to prove the value of its technology in complex enterprise environments.

Despite its high valuation, the reality that OpenAI is not yet profitable is pushing it to actively develop the enterprise market. Applying AI capabilities to specific industries and solving concrete business pain points is seen as the key path to revenue growth and long-term development. The project shows that OpenAI’s ambitions go far beyond general-purpose chatbots; it seeks to develop professional-grade AI tools that deeply embed into enterprise workflows, securing a place in the global business landscape.

Risk Warning and DisclaimerThe market has risks, and investment needs to be cautious. This article does not constitute personal investment advice, nor does it consider the specific investment objectives, financial situation, or needs of any individual user. Users should consider whether any opinions, views, or conclusions in this article are appropriate to their particular circumstances. Investments based on this information are at their own risk. ```