Goldman Sachs is tapping Anthropic’s AI model to automate accounting, compliance roles

Goldman Sachs is making significant strides in the integration of artificial intelligence by partnering with the startup Anthropic to develop AI agents designed to streamline various functions within the bank. According to Marco Argenti, the bank's chief information officer, the collaboration has been ongoing for six months and involves the co-development of autonomous agents focused on areas such as accounting for trades and transactions, as well as client vetting and onboarding processes. Argenti explained that the firm is currently in the initial phases of utilizing Anthropic's Claude model to enhance efficiency in these critical operations. He expressed optimism about launching these AI agents in the near future, although he did not specify an exact timeline. "Imagine these agents as digital coworkers assisting in complex and process-heavy tasks across the organization," he stated. In a recent statement, Goldman Sachs CEO David Solomon highlighted the bank's commitment to a multi-year transformation centered around generative AI, a technology that gained traction following the debut of OpenAI's ChatGPT in late 2022. Despite the current surge in revenue from trading and advisory services, Solomon noted that the bank aims to manage headcount growth during this transition. The news comes at a time when Anthropic, co-founded by a former OpenAI executive, has seen its model updates trigger volatility among software firms and their financiers, prompting investors to reassess potential winners and losers in the AI landscape. Goldman Sachs has already begun experimenting with an autonomous AI coder named Devin, which is now widely accessible to its engineering teams. However, the bank quickly recognized that Anthropic's AI capabilities extend beyond coding. Argenti remarked on the impressive performance of Claude, questioning whether its success in coding was due to the unique nature of programming or its ability to logically navigate complex problems. He noted the model's unexpected proficiency in areas like accounting and compliance, which require extensive data analysis and rule application. As Goldman Sachs continues to explore the potential of AI, Argenti envisions that the automation capabilities could expedite client onboarding and enhance the resolution of issues related to trade reconciliation and other accounting tasks. Future developments may also include AI agents for functions such as employee monitoring or creating investment banking pitchbooks. While the bank employs a substantial workforce in compliance and accounting roles, Argenti emphasized that it is too early to draw conclusions about potential job losses due to AI implementation. However, he acknowledged that advancements in AI could reduce reliance on third-party service providers. "It's a balancing act," Argenti stated. "Our current philosophy is to enhance capacity, enabling us to operate more efficiently, ultimately leading to a superior client experience and increased business opportunities."