Goldman Sachs Uses AI Tools to Shorten Engineering Idea-to-Prototype Cycle

Goldman Sachs can see exactly how much each of its employees uses AI but chooses not to focus on that metric alone to gauge productivity. Marco Argenti, chief information officer at Goldman Sachs, instead measures how quickly his engineering teams can move from coming up with an idea to actually executing on it. A productive team’s work backlog starts shrinking rapidly under this approach.

AI tools have helped Goldman Sachs employees go from creating PowerPoints for their ideas to creating prototypes that can be adjusted in real time based on feedback. “There’s zero time between idea and prototype. You kind of ‘3D print’ software,” Argenti said.

He oversees 12,000 engineers at the investment bank. In 2024, Goldman Sachs launched its GS AI Platform. The platform incorporated large language models from OpenAI and Google with a layer of security to protect the company’s private data.

The bank also launched an in-house version of ChatGPT and a system called “Legend” that lets employees search the bank’s extensive files using natural language. Argenti rejects the idea that simply counting AI interactions reveals whether the technology improves output. “It would be like looking at only one player on the field,” he said.

“Fine, this player is doing more movements, but why am I not scoring more goals? ” The dominant sentiment among Goldman Sachs employees is really a sense of empowerment. People feel almost liberated regarding AI at Goldman Sachs.

A few weeks or months ago there was skepticism and fear, but Argenti correlates that to people who were not really using it. Google has started evaluating some software engineers based on their usage of AI tools. Managers have discretion on whether to take AI tool usage into account for evaluations at Google, according to a Google spokesperson.