On May 5, Anthropic announced ten agent templates for financial services workflows, a Moody’s MCP app embedding credit data on more than 600 million companies, and expanded Microsoft 365 add-ins covering Excel, PowerPoint, and Word.1 Dario Amodei made the announcement alongside JPMorgan CEO Jamie Dimon. The combination, prebuilt agents, proprietary data as native tooling, and a productivity suite foothold, is a direct bid for the workflows currently owned by Bloomberg terminals and Microsoft Copilot.
Ten Agents, Two Tracks
The templates split evenly between client-facing and back-office work. The research and client coverage track includes a pitch builder that assembles target lists, runs comparables, and drafts pitchbooks; a meeting preparer that briefs teams ahead of client calls; an earnings reviewer that reads transcripts and filings, updates models, and flags thesis changes; a model builder for creating and maintaining financial models from filings and data feeds; and a market researcher that tracks sector developments across news and filings.
The finance and operations track covers a valuation reviewer that checks valuations against comparables and methodology; a general ledger reconciler; a month-end closer that runs the close checklist and produces close reports; a statement auditor that reviews financials for consistency and audit readiness; and a KYC screener that assembles entity files, reviews source documents, and packages escalations.
The Register’s coverage describes each template as a reference architecture bundling skills, connectors, and subagents.2 They ship as plugins for Claude Cowork and Code, and as cookbooks for Managed Agents, which supports multi-hour tasks like deal closings with full audit logs.
Claude Opus 4.7 scored 64.37% on the Vals AI Finance Agent benchmark, per The Register;2 independent replication of that number is not yet reported as of this writing.
Moody’s as an MCP App
The Moody’s integration is not a standard connector. Anthropic’s blog post describes it as an MCP app: Moody’s own tools embedded directly inside Claude, bringing proprietary credit ratings and data on more than 600 million public and private companies for compliance, credit analysis, and business development.1 The distinction matters technically. An MCP app means the data source exposes its own interface to the model rather than being piped through an intermediary.
Eight additional connectors also shipped at launch: Dun & Bradstreet, Fiscal AI, Financial Modeling Prep, Guidepoint, IBISWorld, SS&C Intralinks, Third Bridge, and Verisk. These sit alongside an existing roster that already included FactSet, S&P Capital IQ, MSCI, PitchBook, Morningstar, Chronograph, LSEG, and Daloopa.
M365: Three Apps Now, Outlook Pending
Claude add-ins are generally available in Excel, PowerPoint, and Word. Outlook support is listed as “coming soon.” WinBuzzer’s coverage frames this accurately: the May 5 update extends add-ins to PowerPoint and Word and queues Outlook, rather than delivering a complete M365 footprint.3 The M365 integration was first announced in November 2025; this is an expansion, not a launch.
For the core analyst workflow, Excel and PowerPoint cover the load-bearing pair: models in Excel, pitchbooks in PowerPoint. Word adds document drafting; Outlook would close the client communication loop.
What This Does to Incumbents
Reuters reported that finance is now Anthropic’s second-largest enterprise revenue vertical, with 40% of its top-50 customers being financial institutions.4 Named customers include Goldman Sachs, Citi, Citadel, BNY, Carlyle, and Mizuho.
Bloomberg Terminal and Refinitiv Eikon are data businesses. Microsoft Copilot is a productivity suite business. Anthropic’s motion is to put the underwriting data inside the model via MCP and run that model inside the productivity suite via M365 add-ins. That collapses two procurement line items into one negotiation, or at minimum creates pricing pressure on both. WinBuzzer noted that Copilot held 38.6% of enterprise AI usage share as of February 2026, against Anthropic’s 5.7%.3 Closing that gap inside financial services specifically, where Anthropic already has the named customer base, is more tractable than closing it across enterprise generally.
At the announcement, Amodei warned that SaaS companies failing to adapt to AI could “go bankrupt, completely go bust,” according to Reuters.4 He included the standard hedge, firms moving early can do better than ever, but the pressure on vendors whose core value is curated financial data, now available as MCP tooling inside the model, is the obvious implication.
The billing model adds a variable. Anthropic recently shifted enterprise customers above 150 seats to per-token pricing,5 which changes the economics of running Managed Agents on multi-hour tasks. A bank deploying the month-end closer or a KYC pipeline on a per-token meter is doing a different cost calculation than one on a flat seat license. The ten templates give teams a starting point; the open question is what each agentic run actually costs once it hits production volume.
Frequently Asked Questions
How does Anthropic’s finance push compare to OpenAI’s enterprise strategy?
OpenAI holds 25.7% enterprise AI usage share (Feb 2026) — over 4x Anthropic’s 5.7% — and has its own financial-services foothold through partnerships with BNY and BBVA, plus a 3-year workflow deal with ServiceNow. Anthropic’s counter-strategy is vertical depth: rather than competing across all enterprise functions, it is narrowing to finance-specific workflows where it already counts 40% of its top-50 customers.
Do the ten agents cover all of Anthropic’s listed financial-services use cases?
No. Anthropic’s solutions page describes five categories — due diligence, benchmarking, financial analysis, memo and pitch deck generation, and portfolio management — but the ten agents map primarily to financial analysis, memo generation, and back-office operations. Portfolio management and benchmarking have no dedicated template. Separately, Forbes reported Anthropic is building a dedicated anti-money-laundering system that falls outside the current agent set.
How fast is Anthropic’s enterprise tooling adoption actually growing?
Anthropic’s tool-use and workflows API went from 0% enterprise usage share in January 2026 to 5.7% in February, per WinBuzzer. Reuters separately reported Q1 revenue growing at an 80x annualized rate. Both figures predate the May 5 finance-agents launch, meaning the current adoption baseline likely understates post-launch momentum.
Does a potential Anthropic IPO change the pricing risk for banks adopting these agents?
Both Anthropic and OpenAI are reported to be considering IPOs later in 2026. A public offering would pressure Anthropic to demonstrate recurring enterprise revenue, making per-token billing models — like the one imposed on customers above 150 seats — structurally attractive to the company and structurally less predictable for banks running multi-hour agentic workflows at production volume.