Goldman Sachs is developing AI for nuanced processes that go beyond deployment for coding and customer service.
For example, AI struggles with investing because it’s difficult to codify the subtle judgments that make a potential investment attractive, Marco Argenti, chief information officer at the $625 billion financial institution, said during the Evident AI Symposium in New York on Oct. 23. The data service provider and AI think tank.
While the return on investment is clear with equities and other assets, “there are many intangible factors in creating an investment thesis that are hard to translate into AI signals,” Argenti said.
AI can decode how an event has affected an investor’s portfolio after the fact, he said, but Goldman is looking for AI to anticipate the potential of certain events occurring and how they could affect the portfolio.
AI, Argenti said, could eventually help:
- Scout market risks;
- Analyze portfolio exposure to geopolitical events; and
- Provide strategic risk recommendations before an investor asks for it.
The $625 billion bank is slowly building this capability and has no timeline for availability, he said.
Efficiency in existing use cases
While Goldman is pushing the limits of AI, it is also realizing gains from its existing uses of tech, Argenti said.
The FI has rolled out its gen AI-driven coding assistant to 12,000 developers and is seeing a 20% productivity boost, he said.
While the cost of human developers is not declining, the cost of AI developers per token — or string of text — is shrinking and will continue to drop, Argenti said.
In another tech move, Goldman this month launched Goldman Sachs 3.0 organization-wide. The AI-driven platform works as a knowledge finding tool, improving client experience and making employees more efficient by embedding AI in back-office processes such as onboarding clients.
Register here for early-bird pricing for the inaugural FinAi Banking Summit 2026, taking place March 2-3 in Denver. View the full event agenda here.
