Goldman Sachs Deploys AI Agents to Replace Accountants and Compliance Staff

For six months, Anthropic engineers have been embedded inside Goldman Sachs offices with a singular mission: teaching AI to do the work of accountants and compliance officers.

The investment banking giant is developing autonomous agents based on Anthropic'sClaude model to automate two critical back-office functions: trade and transaction accounting, and client vetting and onboarding, according to Marco Argenti, Goldman's chief information officer.

"Think of it as a digital co-worker for many of the professions within the firm that are scaled, are complex and very process intensive," Argenti told CNBC exclusively.

Beyond Code: AI's Surprising Versatility

Goldman initially tested an autonomous AI coder called Devin, which is now widely available to the bank's engineers. But executives were "surprised" at how capable Claude proved to be at tasks far beyond programming.

"Claude is really good at coding," Argenti explained. "Is that because coding is kind of special, or is it about the model's ability to reason through complex problems, step by step, applying logic?"

The answer became clear in accounting and compliance work—areas that demand parsing massive amounts of data and documents while applying nuanced rules and judgment. Claude excelled at these traditionally human-dominated tasks, opening executives' eyes to AI's broader potential.

The Efficiency Promise vs. Job Reality

Goldman CEO David Solomon announced in October a multi-year plan to reorganize the bank around generative AI, explicitly stating the firm would "constrain headcount growth" during the overhaul.

Yet Argenti maintains it's "premature" to expect near-term job losses for the thousands of employees working in compliance and accounting functions where AI agents will soon operate.

The bank's philosophy, he says, is "injecting capacity"—doing things faster to improve client experience and generate more business. However, he acknowledges that third-party providers currently used by Goldman could face displacement as AI technology matures.

"It's always a tradeoff," Argenti said, revealing the delicate balance between efficiency gains and workforce implications.

A Broader Financial Services Shift

Goldman's move signals more than internal innovation—it represents a potential inflection point for the entire financial services industry. As investment banks experience surging revenue from trading and advisory activities, the race to automate back-office operations intensifies.

The bank plans to expand AI agents to employee surveillance and investment banking pitchbook creation next. With competitors watching closely, Goldman's success could trigger industry-wide adoption of similar automation technologies.

This development comes as model updates from Anthropic have sparked sharp selloffs among software firms, with investors scrambling to identify winners and losers in the AI transformation.

The Human Cost of Digital Efficiency

While Goldman emphasizes efficiency over elimination, the language reveals underlying intentions. "Constraining headcount growth" is corporate speak for doing more work with fewer people—even if existing employees keep their jobs.

For financial professionals, the message is clear: adapt or risk obsolescence. The question isn't whether AI will transform finance, but how quickly and extensively.