Goldman Sachs Lets AI Agents Do Accounting and Compliance Work

The initiative reflects the rapid adoption of agentic AI and the broader trends in corporate finance where CFOs and firms are experimenting with productivity-boosting platforms while managing risk and control.

Goldman’s chief information officer, Marco Argenti, told CNBC that Goldman has spent six months embedding Anthropic engineers within its technology teams to co-develop AI agents capable of performing complex, rule-based tasks that extend beyond simple coding or drafting.

These agents are being tested on transaction reconciliation, trade accounting, client vetting and onboarding processes, traditionally labor-intensive jobs that have resisted automation for decades because they require processing large volumes of data against strict regulatory frameworks.

Goldman’s embrace of agentic AI comes amid a broader push toward automation within the finance industry. CEO David Solomon has previously highlighted generative AI as central to a multiyear strategy to control headcount growth and accelerate internal workflows.

Argenti noted that early work with a pilot coding assistant led engineers to realize Claude’s reasoning abilities were strong enough to handle more advanced financial tasks, prompting a move toward deeper use cases.

This deployment is being framed internally as one where AI agents function as “digital colleagues” rather than replacements for human workers, though market reactions hint at larger tension points.

A sell-off in technology and financial services stocks earlier this week, which wiped out billions in market value after Anthropic released a new automation tool, underscores investor concerns that AI could disrupt legacy software vendors and accelerate labor substitution across industries.

CFOs Expand AI Use in Structured Finance Tasks

Goldman’s initiative illustrates how agentic AI is beginning to be used in production across large financial institutions, but it is far from the only strategic push toward autonomous systems. In the corporate world, firms are investing in internal AI platforms that integrate agentic capabilities directly into daily workflow systems.

A case in point is Citi’s rollout of Stylus Workspaces, a platform designed to streamline complex multi-step tasks across applications and data sources, consolidating manual workflows that previously required multiple tools and human handoffs.

Citi’s strategy highlights a broader trend in enterprise AI: firms are choosing to build internal agentic layers rather than rely solely on external products, allowing them to retain control over sensitive financial data and compliance logic while boosting productivity. These platforms can reduce the friction employees face when switching between legacy systems, automate routine work and elevate human capacity for higher-value tasks.

Large numbers of CFOs are already using AI across finance functions, but they are carefully sequencing adoption based on risk and control. According to a December PYMNTS Intelligence report, in structured, rules-based areas like working capital monitoring, cash flow tracking and compliance oversight, 45% of CFOs report using AI tools today, the highest penetration of the technology in any discrete finance domain.

CFOs largely view AI as a visibility and advisory. According to the same PYMNTS Intelligence study, 52% would allow AI to recommend adjustments to liquidity and payment timing, but human oversight is still considered essential in high-risk areas, particularly where cross-system coordination is involved.

Another PYMNTS Intelligence report found that nearly 7% of CFOs have already deployed agentic AI in live finance workflows, while another 5% are running pilots. Even among those not yet deploying the technology, appetite is rising: seven in 10 enterprise CFOs said they are very or extremely interested in using agentic AI for financial planning and analysis, while 68% expressed high interest in applying it to financial reporting and 63% to cost management and working capital optimization.