OpenAI is using more than 100 former investment bankers to train its artificial intelligence how to build financial models, trying to replace the time-consuming and tedious work that junior bankers can do throughout the industry. Documents show that the team includes former employees of JPMorgan Chase, Morgan Stanley and Goldman Sachs Group, and is affiliated with a confidential project code-named "Mercury" within OpenAI.
Participants are paid $150 an hour to write tips and build financial models for a variety of transactions, including restructurings and initial public offerings, a person familiar with the matter said. OpenAI is also offering participants early access to the artificial intelligence it is building to replace tasks in entry-level positions in investment banks.
The project highlights how OpenAI, led by Sam Altman, is working to make its powerful AI technology more practical for companies in consulting, finance, law and technology. The world's largest startup has yet to turn a profit despite being valued at $500 billion earlier this month.
An OpenAI spokesperson said the company works with experts in various fields "to improve and evaluate the capabilities of our models in different domains. Experts are recruited, managed and compensated by third-party vendors."
Investment banking analysts often work more than 80 hours a week while deals are in progress, using Microsoft Excel to build detailed models for mergers and acquisitions and leveraged buyouts. They also often receive instructions from their superiors to modify PowerPoint slides, and then continue to modify them again and again. This culture has given rise to the Wall Street meme of "pls fix."
There are already many startups looking to enter this field and provide banks with AI that can help handle these tasks. While analysts have long complained about boring jobs, the rise of AI is now raising concerns about job security.
According to the people familiar with the matter, the process of applying to participate in the Mercury project involves almost no human interaction. The person requested anonymity discussing non-public information.
The first step is a roughly 20-minute interview with an AI chatbot, which asks questions based on the applicant's resume. The second stage tests the candidate's knowledge of financial statements. The final stage is modeling testing.
The person said the work is quite flexible and outsourcers are expected to submit a model every week. Work instructions include writing prompts using concise wording and then executing the model. After receiving feedback from reviewers, they need to fix the issues before their work is finally plugged into OpenAI's system.
=Documents show that the "Mercury" project has so far attracted participants who have worked at many Wall Street institutions, including those who have worked at Bowen Company, Mubadala Investment Co. , Evercore Inc. and KKR & Co. people. Several current MBA students studying at Harvard and MIT are also participating in the project.
Participants need to create the model in Excel and follow industry specifications for model format, including things like margin size and setting percentages in italics.