The firms selling picks and shovels to Wall Street just became the miners themselves.
The Summary
- Anthropic launched 10 AI agents targeting finance workflows — pitch decks, financial models, meeting prep, market research
- Finance is already Anthropic's second-largest vertical by revenue; 40% of their top 50 customers are banks
- They're entering a market where clients like JPMorgan and Goldman already built internal tools, and startups like $2B Rogo serve 250+ finance shops
- The infrastructure provider is now competing with its own customers
The Signal
Anthropic isn't expanding into finance. They're already there. Finance represents their second-largest enterprise revenue segment after tech, and 40% of their top 50 customers are financial institutions. Tuesday's launch of 10 finance-specific agents isn't a pivot — it's a land grab in territory they've been quietly cultivating.
The agent suite targets the work that defines junior banker life: the model builder creates financial models from SEC filings and analyst notes, the pitch builder drafts client presentations, others handle meeting prep and market research. This is the grunt work that keeps analysts at their desks until 2am, the tasks that consume 60-70% of their time but generate maybe 20% of the value.
"Financial services generates more enterprise revenue for Anthropic than any vertical except tech itself."
Here's the conflict: Anthropic's biggest finance customers have spent the last two years building exactly these capabilities in-house. JPMorgan rolled out AI assistants across large portions of its workforce. Goldman and Morgan Stanley did the same. They licensed Claude, integrated it into their systems, trained it on their data, wrapped it in compliance guardrails, and deployed it to thousands of employees.
Now the company that sold them the underlying model is selling pre-packaged agents that do the same work.
What makes this different from typical platform risk:
- Banks didn't just use Claude as infrastructure — they built strategic moats around it
- The in-house tools are trained on proprietary deal data, client histories, regulatory constraints
- Anthropic's agents are generic by necessity; they can't access the same institutional knowledge
- But they're also turnkey, don't require an AI team to maintain, and work day one
The startup ecosystem saw this coming. Rogo raised money at a $2 billion valuation specifically to build finance agents before the foundation model companies got there. They serve 250+ clients with tools that draft pitch decks, build models, and prep meetings — the exact capability set Anthropic just launched. Former investment bankers founded Rogo precisely because they knew the window for finance-specific agents would close once OpenAI or Anthropic noticed how much banks were spending.
That window just closed. Or rather, it got crowded. The race now is whether vertical-specific startups can build enough institutional depth and network effects to survive commoditization from below, and whether banks' in-house tools can maintain their advantage over plug-and-play alternatives that don't require an AI engineering team.
The Implication
If you're building agents for a specific industry, your moat isn't the agent itself — it's the data, integrations, and institutional knowledge the foundation model companies can't replicate. Banks with mature in-house AI programs have that moat. Startups like Rogo are racing to build it through client relationships and proprietary workflows.
Everyone else is about to compete with Anthropic's agents on price. That's not a race you win. The question for Wall Street now: do you keep building custom agents in-house, or do you shift those teams to problems Anthropic can't solve — the messy, regulated, institution-specific work that still requires humans in the loop?