NEW YORK, Feb 8, 2026, 08:25 EST
Goldman Sachs is teaming up with AI startup Anthropic to develop “AI agents” designed to automate internal banking tasks like trade and transaction accounting, along with client due diligence—the process banks use to verify customers, CNBC reported Friday. Goldman Sachs later confirmed the report, Reuters noted. Reuters
This shift is significant because it moves beyond simple chat assistants to software capable of handling multi-step tasks independently — something banks have long approached cautiously in tightly regulated areas. These AI agents are currently being trialed for transaction reconciliation, trade accounting, client vetting, and onboarding—tasks that, according to PYMNTS, “have resisted automation for decades” due to strict regulatory demands. Pymnts
The timing is tricky for anything billed as “AI that does jobs.” Reuters noted that a selloff hit software and data-analytics stocks after Anthropic’s Claude rolled out a new plug-in. Strategist Carlota Estragues Lopez pointed out that “headlines that would have pushed shares to fresh highs during the peak of AI optimism are now being interpreted far more cautiously by investors.” Reuters
Goldman’s Chief Information Officer, Marco Argenti, told CNBC the firm is “in the early stages” and expects to roll out agents “soon,” though he didn’t specify a timeline, Sharecast reported. He dismissed concerns about job cuts as “premature,” but noted the bank might reduce reliance on some third-party providers as the technology evolves. Sharecast
Argenti framed the concept around speed rather than spectacle. “Picture it as a digital co-worker handling roles within the firm that are scaled, complex, and heavily process-driven,” he told Observer. The report also noted Argenti mentioned future uses could include employee monitoring and crafting investment-banking pitchbooks. Both Goldman and Anthropic declined to comment when approached by Observer. Observer
Beyond Goldman, banks are figuring out which automation moves make sense first — and how to prove they actually deliver. Alexandra Mousavizadeh, cofounder and co-CEO of Evident, told Business Insider that some AI functions have become “table stakes.” Still, AI agents remain in early stages, particularly on the external front where humans likely stay involved as banks iron out guardrails. Businessinsider
Goldman’s shift toward AI started some time ago. A June 2025 Reuters report revealed the bank rolled out its GS AI Assistant across the firm, with roughly 10,000 employees already on board, according to an internal memo shared by Argenti.
Rivals are developing their own internal platforms and workflow tools, all aiming at similar goals, despite different labels. In September 2025, Citi announced an upgrade to its proprietary Citi Stylus Workspaces, adding “Agentic AI” and linking it with select internal systems. Back in June 2024, Morgan Stanley rolled out “Debrief,” an OpenAI-powered tool that generates meeting notes and drafts follow-up messages—client consent required. JPMorgan revealed in 2025 that its LLM Suite roadmap will include AI agents capable of executing multiple steps to achieve a goal. Citigroup
But the risks aren’t just hypothetical. In December, Britain’s Financial Conduct Authority flagged “agentic AI” for posing fresh dangers due to its capacity to act “at pace.” It pointed out that autonomy and speed could amplify governance and stability issues, especially when multiple agents engage with one another. Reuters
Goldman hasn’t announced a public launch date for its agents yet. The initial challenge will be the dull stuff regulators care about: audit trails, controls, and keeping the software reliable when things get complicated. If the agents stumble on edge cases, they won’t cut down on time—they’ll just shift the bottleneck to fixing errors.
