Goldman Sachs just hired its first AI coder

Goldman Sachs just hired a coder who doesn’t sleep, eat, or cash a paycheck. The investment bank announced Friday that it’s testing Devin, an AI software engineer built by London-based startup Cognition, with plans to deploy the autonomous program across its tech teams, according to Goldman’s chief information officer, Marco Argenti.

“We’re going to start augmenting our workforce with Devin, which is going to be like our new employee who’s going to start doing stuff on the behalf of our developers,” Argenti told CNBC. “Initially, we will have hundreds of Devins [and] that might go into the thousands, depending on the use cases.”

Devin rose to prominence last year when Cognition claimed it had created the world’s first AI software engineer. Demo videos showed the system working end to end as a full-stack developer that is capable of completing multistep coding tasks with very little human input. Devin’s “hire” signals more than just an experiment; Goldman has 12,000 developers on staff. 

According to Argenti, the bank plans to scale Devin “into the thousands” — potentially transforming productivity by a factor of three to four (compared with previous-generation assistants) for software teams. Devin will focus on the kind of routine jobs engineers often dread — such as updating legacy code — under the supervision of human developers.

Devin was built by Cognition, a startup founded in late 2023 that is already valued at nearly $4 billion. The company is backed by Palantir’s Peter Thiel and Joe Lonsdale, among others. While Goldman is using Devin, it doesn’t own a stake in the company, a source familiar with the bank’s investments told CNBC.

This move comes amid Goldman’s rollout of its GS AI Assistant, a firm-wide generative AI platform already in use across investment banking, wealth management, and research roles. The platform works with models from OpenAI, Google’s Gemini, and others to power tools for summarizing documents, drafting analyses, and generating production software code.

But Goldman’s embrace of AI coding tools raises some big questions. 

The move comes as AI continues to reshape white-collar work across industries. But rather than using generative AI, Devin represents a new class of “agentic AI” — systems designed to work autonomously, not just to assist with emails or research but to execute entire workflows from start to finish.

The potential implications for labor are already sparking concern. Bloomberg Intelligence said in January that banks could cut as many as 200,000 jobs globally over the next three to five years due to AI adoption. Argenti said Goldman sees a future where humans and AIs work in tandem, describing it as a “hybrid workforce.” 

“It’s really about people and AIs working side by side,” Argenti said. “Engineers are going to be expected to have the ability to really describe problems in a coherent way and turn it into prompts … and then be able to supervise the work of those agents.”

But while the official messaging might emphasize augmentation over replacement, Goldman’s president, John Waldron, in late May, cited 20% productivity gains among developer teams using AI tools, hinting at leaner future headcounts.

Not all roles are created equal. MarketWatch reporting warned that traditional analyst paths — the entry‑level conveyor belt — could shrink dramatically. Still, jobs that demand sector expertise, client relationships, and high‑stakes judgment are expected to remain resilient, even as routine coding and research tasks become increasingly automated.

Argenti told CNBC of Devin: “Those models are basically just as good as any developer; it’s really cool.” Whether that’s terrifying or exciting probably depends on at which side of the keyboard you sit.