Goldman Sachs has started testing Devin, an autonomous coding agent created by Cognition Labs, treating it as a new “employee” to work alongside its ~12,000 human developers. Initially, the bank will roll out hundreds of instances, with plans to scale to thousands depending on use case success
Why It Matters
CIO Marco Argenti explained that Devin will handle repetitive or “drudgery” coding tasks—like updating legacy code—under human oversight. He calls this a hybrid workforce, where AI augmentation boosts developer productivity rather than replacing jobs
How Devin Works
Devin is considered a leap in agentic AI—able to autonomously write, test, debug, and sometimes deploy code on its own. It can navigate code editors, file systems, and online resources to execute multi-step tasks with minimal human input. According to Cognition, version 2.1 excels in large, context-rich codebases
Real-World Feedback & Caution
While Goldman’s move signals confidence, independent tests show Devin still struggles with complex tasks—succeeding in only a fraction of trial assignments . Still, Goldman emphasizes the need for human supervision and oversight.
Implications for the Developer Workforce
- Productivity gains: Argenti suggests Devin could offer 3–4× speed improvements for some tasks
- Workforce shift: Engineers will increasingly focus on prompt design, code review, and systems oversight.
- Job market effects: While Devin may replace certain entry-level or routine roles, Goldman frames it as augmenting—not eliminating—the human workforce The Times of India.
Bottom Line
Goldman Sachs’ rollout of Devin, an AI-driven “software engineer,” marks a bold leap in enterprise AI use. By blending autonomous agents with skilled oversight, Goldman aims to redefine software development productivity—offering a blueprint for “hybrid” tech teams across industries.
