Goldman Sachs has integrated engineers from Anthropic for a six-month period to develop automated systems capable of handling time-consuming, high-volume back-office operations.

The bank anticipates that this technology will enhance operational efficiency. While no immediate layoffs are planned, artificial intelligence is expected to accelerate business processes and help control future headcount growth.

Anthropic's model has demonstrated strong performance in areas beyond code generation, surprising Goldman Sachs executives. This confirms that AI can manage complex, rules-based tasks such as accounting and compliance.

Goldman Sachs' Chief Technology Officer exclusively revealed to CNBC that the bank is collaborating with AI startup Anthropic to build AI agents aimed at automating more roles within the institution.

Chief Information Officer Marco Argenti stated that over the past six months, Anthropic engineers have been embedded within Goldman Sachs to jointly develop these automation agents. The current focus is on two specific areas: trade bookkeeping, and client qualification checks and account opening processes.

Argenti noted that the agents, developed using Anthropic's Claude model, are still in the "early stages of development." They are expected to significantly reduce processing times for these core functions. Goldman Sachs plans to roll out the product "as soon as possible," though a specific launch date was not provided.

He described these agents as "digital colleagues" for various roles at Goldman Sachs, capable of handling scalable, complex processes that heavily rely on standardized operations.

In October 2025, Goldman Sachs CEO David Solomon announced a multi-year plan to reorganize the business around generative AI, a trend accelerated by the launch of OpenAI's ChatGPT in late 2022. Solomon added that despite significant revenue growth in the bank's trading and advisory businesses, the firm aims to "control headcount growth" during this transformation.

Anthropic, founded by former OpenAI executives, has seen its model upgrades significantly impact the stock prices of software companies and related credit agencies, as investors bet on potential winners and losers in the AI wave. The news of this partnership with a top-tier global investment bank emerges in this context.

Goldman Sachs began testing an automated AI programming tool named Devin in 2025, which is now available to all engineers within the bank. Argenti mentioned that the bank quickly discovered that Anthropic's AI models could also be applied to other business segments.

"Claude performs exceptionally well at writing code," Argenti said. "The question was whether this was unique to programming, or if the model possessed the ability to reason through complex problems step-by-step and apply logic to solve them."

He stated that Claude's powerful capabilities extended to non-programming tasks, which was a "surprise" to Goldman Sachs, particularly in accounting and compliance. These areas require parsing vast amounts of data and documents, as well as making professional judgments based on rules, tasks which Claude handled adeptly.

Argenti stated that the consensus within Goldman Sachs is now that "other business areas of the bank have the potential to achieve automation levels and results as impressive as those in the programming segment."

He indicated that once deployed, these agents will yield significant benefits: client onboarding will be faster, and the efficiency of trade reconciliation and other accounting-related issue resolution will be greatly improved.

Argenti revealed that the next phase involves developing agents for monitoring employee conduct and creating investment banking pitch materials.

Although Goldman Sachs' accounting and compliance departments currently employ thousands of staff, and AI agents will soon be introduced to these roles, Argenti believes it is "premature" to conclude that the technology will lead to job losses for these employees.

However, he suggested that as AI technology matures, Goldman Sachs might reduce its reliance on some current third-party service providers.

"Everything is a trade-off," Argenti commented. "Our current philosophy is to use AI to enhance our processing capabilities. In most cases, this makes us more efficient, leading to a better client experience and allowing us to win more business."