Anthropic launches ten finance AI agents and Office add-ins

TL;DR:

  • Anthropic has released ten “ready-to-run” agent templates for financial services, covering pitchbook building, KYC screening, month-end close, valuation review, financial-model building, statement audit and earnings reviews. Each ships as a plugin in Claude Cowork and Claude Code, and as a cookbook for Claude Managed Agents.
  • Claude is now embedded in Microsoft Excel, PowerPoint and Word via add-ins (Outlook coming soon), with new connectors for FactSet, S&P Capital IQ, MSCI, PitchBook, Moody’s, IBISWorld, Dun & Bradstreet, Verisk and others.
  • Resultsense view: this is the most concrete enterprise vertical play any frontier lab has shipped to date. The architecture — agent templates, Microsoft-365 surface, governed connectors to industry data — is the pattern UK financial-services CIOs should expect competitors to follow within two quarters.

The release positions Claude Opus 4.7 as Anthropic’s “state-of-the-art” finance model, citing a 64.37% score on Vals AI’s Finance Agent benchmark. Carlyle, Walleye Capital, BMO and Amalgamated Bank are among the named adopters.

What the agents do

The ten templates split into two groups. Research and client coverage: Pitch builder, Meeting preparer, Earnings reviewer, Model builder and Market researcher. Finance and operations: Valuation reviewer, General ledger reconciler, Month-end closer, Statement auditor and KYC screener. Each agent packages skills (instructions and domain knowledge), connectors (governed access to source data) and subagents (specialised models for sub-tasks like comparables selection or methodology checks).

The templates can run two ways: as a plugin alongside the analyst on the desktop, or as a Claude Managed Agent running autonomously on the Claude Platform with long-running sessions, per-tool permissions, managed credential vaults and an audit log in the Claude Console.

The Microsoft 365 surface

Claude add-ins for Excel, PowerPoint and Word are generally available; Outlook is coming soon. In Excel, Claude builds and audits financial models from filings and data feeds, runs sensitivity analyses and audits formulas across linked workbooks. In PowerPoint, decks update automatically when underlying numbers change. In Word, Claude edits credit memos against a firm’s templates. Context carries between applications — an analyst who starts a model in Excel does not have to re-explain it when the work moves to PowerPoint.

The Outlook integration is positioned as a “chief of staff”: triaging inbox, arranging meetings, drafting responses in the user’s voice. That moves Claude into territory Microsoft’s own Copilot has owned to date.

The data ecosystem

New connectors include Dun & Bradstreet (verified business identity), Fiscal AI (fundamentals), Financial Modeling Prep, Guidepoint (100,000+ compliance-reviewed expert interview transcripts), IBISWorld (industry data), SS&C IntraLinks (deal-room access), Third Bridge (expert interviews) and Verisk (insurance underwriting). Moody’s has launched a separate MCP app providing credit ratings and data on more than 600m public and private companies. Existing connectors already cover FactSet, S&P Capital IQ, MSCI, PitchBook, Morningstar, Chronograph, LSEG and Daloopa.

UK relevance

Three points for UK financial-services buyers. First, the Microsoft 365 surface bypasses much of the desktop-procurement friction UK banks usually face deploying new AI tools — Claude lives where the analyst already works. Second, the connector list reads like the data spine of the City of London — FactSet, S&P, MSCI, PitchBook, Moody’s, LSEG. UK funds that already pay for these services will get materially less greenfield-procurement complexity than peers building bespoke. Third, the architecture sits cleanly inside FCA SYSC outsourcing guidelines: governed access controls, per-tool permissions, full audit log, human review before action — all the surfaces UK regulators have signalled they want to see.

Looking forward

The single biggest open question is auditability. UK financial regulators have been signalling for two quarters that agentic AI in financial-services contexts will need an audit trail granular enough to reconstruct any individual decision. Anthropic’s Console audit log is the right shape; whether it satisfies FCA examination standards under SS1/23 and SYSC 8 remains untested. UK firms running early pilots should prioritise the audit-trail evaluation before scope-creeping further into autonomous deployments.

Share this article