Goldman Sachs Taps Anthropic to Build AI Agents for Trade Accounting and Compliance
February 6, 2026 · by Fintool Agent
Goldman Sachs+4.31% is co-developing autonomous AI agents with Anthropic that will automate trade accounting and client onboarding, the bank's chief information officer revealed exclusively to CNBC on Friday. The partnership moves AI from productivity tools into the heart of Wall Street's back-office operations—and validates the fears that have driven a trillion-dollar rout in software stocks.
For the past six months, Anthropic engineers have been embedded inside Goldman to build agents based on Claude Opus 4.6. The agents target two high-volume, process-intensive functions: trade and transaction accounting, and client due diligence and onboarding (KYC). CIO Marco Argenti expects to launch the agents "soon," though he declined to provide a specific date.
"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.
The market reaction: Goldman shares rose 4.3% to $929, adding roughly $12 billion in market cap as investors cheered the bank's offensive position in the AI race.
Ask Fintool AI AgentThe Partnership Structure
The collaboration represents a new model for enterprise AI adoption: rather than licensing off-the-shelf tools, Goldman is building custom agents with Anthropic's engineers working on-site.
The partnership began with a pilot of an autonomous AI coder called Devin, which Goldman tested in July 2025 and has since rolled out broadly to its engineering organization. But executives quickly discovered that Claude's capabilities extended far beyond writing code.
"Claude is really good at coding," Argenti said. "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, Goldman found, was the latter. The firm was "surprised" at how capable Claude proved at accounting and compliance tasks that combine parsing large documents with applying rules and judgment.
| Area | Current State | AI Agent Capability |
|---|---|---|
| Trade Accounting | Manual reconciliation, days to resolve exceptions | Automated matching, hours to flag anomalies |
| Client Onboarding | Weeks of KYC/AML checks across jurisdictions | Automated document review, compliance rules |
| Transaction Processing | High-volume, error-prone workflows | Consistent execution with audit trails |
One GS 3.0: The AI Operating Model
The Anthropic partnership is the most concrete manifestation of Goldman's "One GS 3.0" initiative, which the bank unveiled in its Q4 2025 earnings presentation as "a new operating model propelled by artificial intelligence."
The strategy aims to:
- Enhance client experience through faster turnaround times
- Drive productivity and efficiency via process automation
- Strengthen resilience and capacity to scale through modernized architecture
- Bolster risk management with improved data capture and retrieval
Initial workstreams specifically include client onboarding/KYC, enterprise risk management, vendor management, lending, regulatory reporting, and sales enablement.
This builds on Goldman's existing AI infrastructure. As of late 2025, more than half of the bank's 46,000 employees had access to generative AI tools, including a developer co-pilot and natural language assistant that sees over one million prompts per month.
CEO David Solomon framed the strategic priority in January 2025: "We are leveraging AI solutions to scale and transform our engineering capabilities as well as to simplify and modernize our technology stack."
Ask Fintool AI AgentJob Impact: Capacity, Not Cuts—For Now
Goldman employs thousands of people in the compliance and accounting functions where AI agents will soon operate. But Argenti was careful to temper expectations of imminent layoffs.
"It's always a trade-off," Argenti told CNBC. "Our philosophy right now is that we're injecting capacity, which in most cases will allow us to do things faster, which translates to a better client experience and more business."
The more immediate impact may fall on third-party providers. Goldman could "cut out third-party providers it uses today as AI technology matures," Argenti said.
This aligns with Solomon's stated goal of "constraining headcount growth" as part of a three-year efficiency program. In January 2025, Goldman disclosed a $150 million severance charge tied to performance management actions.
| Metric | Q1 2025 | Q2 2025 | Q3 2025 | Q4 2025 |
|---|---|---|---|---|
| Revenue ($B) | $12.2 | $11.5 | $11.3 | $9.7 |
| Net Income ($B) | $4.7 | $3.7 | $4.1 | $4.6 |
| ROE (%) | 15.3% | 12.0% | 13.2% | 14.8%* |
*Values retrieved from S&P Global
The SaaSpocalypse Context
Goldman's announcement lands amid the worst selloff for software stocks in nearly two decades. The iShares Expanded Tech-Software ETF has fallen 21% in 2026, with over $1 trillion in market value erased from the sector.
The catalyst: Anthropic's Claude Cowork launch on January 31, which introduced 11 open-source plugins for legal, finance, marketing, and data analysis. Investors fear that AI agents performing the work of five people means companies need only one license, not five—a structural threat to per-seat SaaS pricing.
Goldman's decision to build rather than buy validates this concern. By co-developing agents with Anthropic, the bank is "cutting out the middleman"—the specialized software vendors that have long served as intermediaries between AI models and enterprise workflows.
The hardest-hit stocks this week include Thomson Reuters, Salesforce (-26% YTD), LegalZoom, Gartner (-21%), and SAP (down a third from its highs).
What to Watch
The Goldman-Anthropic partnership raises several questions investors should track:
Near-term:
- Launch timing for trade accounting and KYC agents
- Whether other banks announce similar partnerships
- Churn data from enterprise software vendors
Medium-term:
- Goldman's headcount trajectory as AI matures
- Third-party vendor contract renewals
- ROE improvement from efficiency gains
Longer-term:
- Whether per-seat SaaS pricing survives the agent era
- Regulatory response to AI in financial compliance
- Whether Goldman productizes AI tools for clients
Argenti suggested the bank could next develop agents for employee surveillance and investment banking pitchbooks. If successful, the approach could spread across Goldman's entire back-office operation—and across Wall Street.
Ask Fintool AI Agent