OpenAI’s New Plan to Automate Wall Street

OpenAI is training its AI to do the grunt work of Wall Street’s youngest bankers, and it’s paying industry veterans handsomely to teach it.

According to documents reviewed by Bloomberg, a new initiative at OpenAI code-named "Project Mercury" has contracted more than 100 former bankers from top firms, including J.P. Morgan, Morgan Stanley, and Goldman Sachs.

Their job? To create financial models for IPOs, restructurings, and buyouts, effectively training AI to automate the very tasks that define the junior banker experience.

This move reveals a clear playbook for how AI labs plan to conquer high-value industries. To understand the strategy, I turned to SmarterX and Marketing AI Institute founder and CEO Paul Roetzer on Episode 176 of The Artificial Intelligence Show.

The "Project Mercury" Playbook

The initiative offers a fascinating glimpse into how frontier AI models are specialized. Participants are reportedly paid $150 an hour to write prompts and build the complex Excel spreadsheet models used for major transactions.

This workflow mirrors the traditional analyst experience: submit a model, get feedback from a reviewer, fix issues, and ship updates. The difference is that the updates are being fed directly into an AI model.

But why banking? According to Roetzer, this is simply the most visible example of a much larger strategy.

“This is the playbook,” he said. “You pre-train the model and then you go in and you fine-tune it to do specific jobs and that's how you automate work.”

Targeting Trillions in Knowledge Work

Roetzer explains that this same fine-tuning process is almost certainly happening in other high-stakes professions.

“I can almost guarantee you they're doing this with lawyers and accountants and consultants and, take your pick,” he said.

The ultimate goal isn't just to create a "copilot" for bankers. It's to capture a piece of the single largest value pool in the modern economy: knowledge work.

“There's $11 trillion in U.S. wages, probably five to six trillion of that is for knowledge workers,” said Roetzer. “The greatest way to build wealth is you automate knowledge work."

While OpenAI says it regularly works with outside experts to improve its models, Project Mercury underscores a new, aggressive push into specific industries.

“This is how the next three to five years goes,” said Roetzer. “You just pick industry at a time, vertical at a time, and you just go train a model to do that work.”