Goldman Sachs Uses Anthropic AI to Drive Business Success

Goldman Sachs is set to deploy Anthropic's Claude AI model in trade accounting and client onboarding — a strategic move that signals a broader shift among major financial institutions toward generative artificial intelligence for operational efficiency. According to a report by American Banker, the initiative targets back-office processes that have historically depended on large teams performing document review, reconciliation, and compliance checks.

Major Banks Are Already Embracing Generative AI

A wave of AI adoption is already underway across the financial sector:

• JPMorgan Chase — Equips employees with a large language model suite for information retrieval and data analysis.
• Bank of America — The Erica assistant handles internal technology and HR queries.
• Citi & Goldman Sachs — Both leverage AI to support developer coding workflows.

The emerging frontier is the application of generative AI to operational tasks — most notably in trade accounting and Know Your Customer (KYC) compliance processes.

⚡ Automating the Edge Cases

Most automatable financial processes are rules-based: collecting data, validating it against internal and external databases, and assembling required documentation. Conventional software has long handled such workflows — but consistently struggles when transactions fall outside predefined parameters.

💬 Marco Argenti, Chief Information Officer, Goldman Sachs:
"If a rules-based system resolves most cases, a small percentage of transactions fall outside defined parameters — which can translate into thousands of individual items at the scale we operate. In KYC compliance, for instance, minor discrepancies or documents nearing expiry can create edge cases that require genuine judgement."

Argenti contends that neural networks are uniquely capable of handling these micro-decisions, applying contextual reasoning where fixed rules are absent or ambiguous. In this model, generative AI augments existing rules-based systems rather than supplanting them — reducing the volume of exceptions requiring manual intervention and shortening overall resolution time.

💻 Goldman's Coding Experience: The Foundation for Broader AI Deployment

Goldman's confidence in expanding AI into operations was built on its prior use of Claude models in software development. Developers use a version of Claude integrated with Cognition's Devin agent to assist with programming — following a clear division of responsibilities:

• Human developers define specifications and regulatory parameters.
• The AI agent produces code and runs tests and validations.
• Human reviewers assess all outputs before deployment.

The outcome: measurably higher developer productivity and faster project completion — with human oversight preserved throughout every stage.

📋 Deploying AI in Trade Accounting & Client Onboarding

For trade accounting and client onboarding, Goldman and Anthropic's teams conducted detailed observations of existing workflows alongside domain experts to identify operational bottlenecks. The Claude-powered agents are now capable of:

• ✓  Reviewing and extracting information from complex documents
• ✓  Determining whether additional documentation is required
• ✓  Assessing corporate ownership structures
• ✓  Triggering further compliance checks when anomalies are flagged

By automating extraction and preliminary assessment, these agents significantly reduce the time analysts spend on comparison work — particularly in document-heavy workflows that previously demanded individual judgement at every step.

💬 Indranil Bandyopadhyay, Principal Analyst, Forrester:
"Reconciliation in trade accounting requires comparing fragmented data across internal ledgers, counterparty confirmations, and bank statements. Claude's ability to process large context windows and follow nuanced instructions makes it ideally suited to these workflows. In client onboarding — parsing passports and corporate registration documents — AI's capacity to extract structured data and flag inconsistencies is a strong fit, materially reducing overall workloads."

Bandyopadhyay emphasizes that accounting and compliance platforms remain the canonical systems of record. Claude operates within the workflow layer — handling extraction and comparison so human analysts can concentrate on exceptions. In regulated environments like banking, this division of labour is precisely where operational value is realised.

🔒 Auditability, Oversight & Hallucination Mitigation

💬 Jonathan Pelosi, Head of Financial Services, Anthropic:
"Claude is trained to surface uncertainty and provide source attribution, creating a full audit trail — directly reducing the impact of hallucinations in compliance-sensitive workflows."

Bandyopadhyay reinforces the critical importance of human oversight and early error detection, urging institutions to design AI systems where anomalies are identified and addressed before they propagate downstream.

Goldman's Argenti challenges the assumption that AI systems are inherently more vulnerable to manipulation than human decision-makers — noting that social engineering targets human psychology, while AI can detect subtle anomalies at scale that would be invisible to individuals. His argument implies the potential for meaningful increases in operational capacity without proportional growth in headcount, even accounting for the known challenges of large-scale AI deployment.

🏢 The Broader Picture: Generative AI Reshaping Banking Operations

In the financial services industry, generative AI is becoming a core operational lever — driving measurable improvements by:

• ▶  Accelerating document processing at scale
• ▶  Reducing exception handling and resolution time
• ▶  Increasing throughput in high-volume compliance workflows

Yet the imperative to retain human oversight — to catch and correct AI errors — means existing systems of record remain indispensable. The technology augments, rather than replaces, the human-in-the-loop model that regulatory environments in banking demand.

Image: "Dreams…" by noahwesley, licensed under CC BY-NC-SA 2.0

Want to learn more about AI and big data from industry leaders? Check out the AI & Big Data Expo — taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading technology events as part of TechEx. Click here for more information.

AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.