Goldman Sachs says it has been working with Anthropic on AI agents to automate more roles at the bank, like trades, transactions, client vetting, and onboarding
CNBCHugh Son
Context & Ripple Effects
Goldman Sachs had already moved from general-purpose productivity tools to workforce augmentation: it launched a companywide AI assistant with roughly 10,000 users and later said it would augment its workforce with Cognition's Devin. The Anthropic work extends that path into banking workflows where transactions, client vetting, onboarding and trading operations are central.
The significance is not simply another employee copilot. Goldman is identifying role-level processes for agents, making its relationship with Anthropic a test of whether a frontier-model provider can be embedded in tightly controlled financial operations.
First-order effects
Goldman Sachs can apply Anthropic-developed agents to specified operational and revenue-adjacent workflows, shifting those teams from performing routine steps to supervising, reviewing and escalating agent work.
Anthropic gains a named bank deployment effort spanning multiple functions, strengthening its position as a supplier of financial-services automation rather than only a general AI assistant provider.
Second-order effects
Goldman's prior companywide AI-assistant rollout creates a user and workflow base from which agent deployments can be evaluated, increasing pressure to connect productivity tools to measurable process automation.
Other banks and AI vendors will be pushed to demonstrate comparable controls for client onboarding, vetting and transaction workflows, where reliability and human oversight matter as much as model capability.
Third-order effects
If deployments move beyond pilots, bank AI adoption may reorganize around embedded, supervised agents for discrete processes rather than standalone chat interfaces—changing how institutions buy software, define operations roles and allocate oversight.
The pattern raises the importance of governance that can document agent actions and handoffs in regulated workflows; the pace of broader automation will depend on whether banks can operate those controls consistently.
The trend: This is part of the shift from enterprise AI copilots toward embedded agents that execute bounded, high-value financial workflows under institutional oversight.
there is something funny about white collar jobs just getting obliterated from the market maybe it's a good thing, we will have less opinions on this site
Last year, Claude worked with Norway's ~$2T sovereign wealth to create a Financial Analysis tool plugged into real-time data (Factset, Morningstar, Pitchbook, Palantir, S&P). Fund's CEO said it led to ~20% (or 213,00 work hours) productivity gains by “automating monitoring of [im…
I think it's totally likely that in 12 years we're hoping our daughter gets into a top performing-arts school because live human performance is well-compensated vs knowledge work
The beginning of labor disruption and “AI workers” tackling key back office roles Beyond coding (where we'll see real impact this year across F500 enterprises) to accounting, contracting, excel modeling & many more...
The worst part of this whole disempowerment saga is that the obnoxious tech bros who claimed that you only needed to automate software in order to automate all work may have been right
Anthropic CEO Dario Amodei warns: “50% of entry-level white-collar jobs could be disrupted within the next 1-5 years.” From law to finance to consulting, he says that AI is capable of handling a wide range of knowledge work. [video]
Anthropic is like The Joker complaining about Gotham City's undrinkable water while dumping barrels of arsenic in the water supply. — The CEO is constantly in the news about how AI will cause mass unemployment and yet at the forefront of selling AI tools to automate jobs. Bril…