goldman sachs is working with artificial intelligence startup Anthropic to develop AI agents that will automate a growing number of roles within banks, the company’s technology chief told CNBC exclusively.
Marco Argenti, Goldman’s chief information officer, said the bank has been working with embedded Anthropic engineers over the past six months to co-develop autonomous agents in at least two specific areas: trades and transaction accounting, and customer vetting and onboarding.
Argenti said the company is in the “early stages” of developing a drug based on Anthropic’s Claude model that disrupts the time it takes for these critical functions. He expects to launch the agency “soon,” but declined to give a specific date.
“Think of this as the digital colleague for many of our in-house professionals who are large, complex, and require a huge number of processes,” he said.
Goldman Sachs CEO David Solomon said in October that the bank is embarking on a multi-year plan to reorganize around generative AI, a technology that has been buzzing since the debut of OpenAI’s ChatGPT in late 2022. Solomon said the bank will aim to “control headcount growth” in the overhaul, even as investment banks like Goldman have seen soaring revenues from trading and advisory activities.
The news from Goldman, one of the world’s leading investment banks, comes as a model update from Anthropic, co-founded by a former OpenAI executive, sparked a sharp selloff among software companies and their credit providers as investors bet on winners and losers in the AI industry.
Last year, Goldman began testing an autonomous AI coder called Devin, which is now widely available to the bank’s engineers. But Argenti said he quickly learned that Anthropic’s AI models worked in other parts of the bank as well.
“Claude is really good at coding,” Argenti said. “Is it because the coding is kind of special, or is it about the model’s ability to reason through complex problems step by step, applying logic?”
Argenti said the company was “surprised” by Claude’s ability to do more than just coding, especially in fields such as accounting and compliance, which require parsing large amounts of data and documents while applying rules and judgment.
The current view within Goldman is that “we can expect the same level of automation and the same level of results in other areas of the company that we’re seeing on the coding side,” he said.
As a result, customers will be onboarded more quickly and issues related to trade adjustments and other accounting issues will be resolved more quickly with the help of agents under development, Argenti said.
Goldman could next develop agents to perform tasks such as monitoring employees and writing pitches for investment banks, he said.
The bank employs thousands of people in its compliance and accounting departments, where AI agents will soon go live, but Argenti said it is “premature” to predict that the technology will lead to job losses for those employees.
Still, as AI technology matures, Goldman may discontinue the third-party providers it currently uses, he said.
“It’s always a tradeoff,” Argenti said. “Our philosophy at the moment is that we are injecting capacity, which in most cases allows us to do things faster, which leads to a better customer experience and more business.”
