OpenAI’s Project Mercury Enlists Ex-Bankers to Train Financial AI Models

OpenAI is quietly recruiting former investment bankers to train its finance-focused AI models under an internal initiative dubbed 'Project Mercury.' These contractors, paid $150 per hour, feed real-world dealmaking data—spanning IPOs, restructurings, and Leveraged buyouts—into the system. The goal: automate the grueling Excel and PowerPoint workflows of junior Wall Street analysts.

The project represents one of OpenAI's most ambitious forays into corporate workflows, targeting industries reliant on complex quantitative analysis. Despite its $500 billion valuation, the startup remains unprofitable, making such experiments critical for proving commercial viability. Former bankers build complete financial models weekly, with each submission reviewed before integration.

Contractors operate through third-party suppliers, and applicants undergo rigorous screening. The initiative highlights AI's encroachment into high-stakes financial domains—a trend that could reshape traditional finance roles while creating new opportunities for crypto-native firms.