Goldman Sachs deploys Anthropic's new AI model for back-office tasks
What's the story
Goldman Sachs is revolutionizing its operations by deploying autonomous artificial intelligence (AI) agents powered by Anthropic's Claude Opus 4.6. The move comes after six months of in-house development and integration with Anthropic engineers. These AI agents are now handling complex back-office tasks such as trade accounting, compliance, and client onboarding—areas that were previously resistant to automation.
Asset management
Claude's capabilities
The deployment of these AI agents involves over 12,000 developers and thousands of back-office staff using Claude's advanced reasoning to handle a portion of the bank's $2.5 trillion in assets under supervision. Unlike traditional chatbots, these autonomous agents leverage Claude 4.6's massive one-million-token context window to process complex financial data in real-time.
Productivity gains
Increase in developer productivity
Goldman Sachs has already seen a significant increase in developer productivity, with AI-driven coding assistants boosting their efficiency by over 20%. This translates into thousands of manual labor hours saved every week. The bank's CIO, Marco Argenti, said the focus has shifted from simple coding assistance to complex financial tasks that require logic and reasoning.
Compliance assurance
AI agents powered by 'constitutional AI'
The AI agents are powered by Anthropic's "constitutional AI" framework, which ensures strict regulatory compliance. This is a major shift in how Wall Street processes high-stakes transactions. By deploying these "digital co-workers," Goldman Sachs hopes to keep rising operational costs in check while handling record-breaking trade volumes. The partnership marks a new era where generative AI goes beyond creative text and into the data-heavy world of global Tier-1 investment banking.
Onboarding efficiency
Reduction in time taken to onboard institutional clients
Goldman Sachs has also seen a 30% reduction in the time taken to onboard new institutional clients by leveraging Claude's logic-based reasoning. The AI agents navigate "Know Your Customer" (KYC) and Anti-Money Laundering (AML) protocols by cross-referencing global databases against internal compliance standards. This data-heavy approach ensures every transaction meets the Federal Reserve's stringent oversight requirements.
Task automation
Internal trials with Claude
During internal trials, Goldman Sachs engineers applied Claude to accounting workflows, compliance reviews, and regulatory documentation. These tasks involve dense rulebooks, fragmented data, and high error sensitivity. According to Argenti, the results surprised the team as Claude was able to interpret policy language, apply conditional logic, and complete multi-step processes that typically require trained professionals.