Goldman Sachs taps Anthropic for AI agents: Chief information officer explains one capability that 'surprised' investment bank - The Times of India

Goldman Sachs has begun deploying fully autonomous AI agents, developed in collaboration with Anthropic, to streamline and automate complex roles within its investment banking operations. These "digital co-workers" are powered by Anthropic's Claude model. According to Marco Argenti, Goldman's Chief Information Officer, the bank spent six months co-developing these AI agents with Anthropic engineers. Initially, Goldman successfully tested AI coding assistants for its engineers before transitioning to Anthropic's Claude models. Argenti highlighted the team's surprise at Claude's robust reasoning capabilities, especially when handling high-stakes financial tasks. He noted that Claude excels not only in coding but also in its ability to logically work through complex problems step-by-step. This move is part of a multi-year reorganization led by CEO David Solomon, aiming to center the bank's operations around generative AI. Goldman Sachs intends to explore the potential of these AI agents to provide similar levels of automation and results in other departments beyond coding, such as employee surveillance or creating investment banking pitchbooks.