Anthropic releases 10 Claude agent templates for finance (pitchbooks, KYC, month-end close), Microsoft 365 add-ins, and 8 new data connectors including Moody’s MCP app.
Key Takeaways
Templates ship as plugins in Claude Cowork and Claude Code, or as Claude Managed Agents with long-running sessions, per-tool permissions, credential vaults, and full audit logs in Claude Console.
Microsoft 365 add-ins for Excel, PowerPoint, and Word are GA; Outlook is coming soon. Context carries automatically across apps so work started in Excel flows to PowerPoint without re-prompting.
Claude Opus 4.7 leads Vals AI Finance Agent benchmark at 64.37%, cited as rationale for pairing with these templates.
New connectors include Dun & Bradstreet, SS&C IntraLinks, Guidepoint, Third Bridge, IBISWorld, Verisk, Fiscal AI, and Financial Modeling Prep; Moody’s ships an MCP app covering 600M+ public and private companies.
Named enterprise adopters include FIS (AML investigations), Carlyle, Walleye Capital (100% Claude Code adoption across 400 staff), and FactSet.
Hacker News Comment Review
Core skepticism centers on hallucination risk in high-stakes numerical workflows; commenters question how firms can verify Claude did not alter numbers, especially in GL reconciliation or financial statements.
Several commenters see this as Anthropic competing directly with the vertical AI startup ecosystem, compressing the moat for any LLM wrapper built on Claude in finance.
A minority of practitioners report real operational use in spend management and receipt-to-GL matching, suggesting narrow, auditable sub-tasks are viable before broader agentic deployment.
Notable Comments
@OkayPhysicist: Reports production success using LLMs to verify receipt-to-GL code matching, with human override logged, as a concrete working pattern.
@motbus3: “The only reason they are doing it is because there are regulation for people but not for machines.”
