Anthropic storms Wall Street: 10 finance agents, Claude inside Excel/PowerPoint, Moody's as a native app
On May 5, 2026, Anthropic unveiled ten ready-to-run agent templates for financial services (pitchbook, KYC, month-end close), full Microsoft 365 integration (Excel, PowerPoint, Word, Outlook) and a Moody's partnership embedding the credit-rating leader directly inside Claude. A frontal attack on a market long locked up by Bloomberg, FactSet and Capital IQ.
On May 5, 2026, Anthropic published its official announcement: ten agent templates for financial services, full Microsoft 365 integration (Excel, PowerPoint, Word, with Outlook in beta), and a data partnership with Moody's that embeds the rating-platform leader directly inside Claude. Fortune called it a "Wall Street assault." PYMNTS and Crypto Briefing described it as the most aggressive vertical play any AI lab has run to date.
It's also a strategic signal: Anthropic is no longer content being a model vendor. The lab is now pushing packaged business workflows that bite directly into Bloomberg Terminal, FactSet, S&P Capital IQ, and Refinitiv territory.
The 10 agent templates: what each one does
Anthropic ships ten preconfigured agents in three forms: Claude Cowork plugin, Claude Code command, and Claude Managed Agents cookbook. A team can deploy an agent in days rather than months.
| Agent | Task covered |
|---|---|
| Pitch Builder | Builds an IBD pitchbook from filings and internal research |
| Meeting Preparer | Prepares a client briefing (positions, deals, news) ahead of a meeting |
| Earnings Reviewer | Analyzes a quarterly report and benchmarks against consensus |
| Model Builder | Builds a financial model (DCF, comps) from filings and data feeds |
| Market Researcher | Compiles a sector thesis from primary and secondary sources |
| KYC Screener | Filters KYC files with compliance checklist and adverse media alerts |
| Valuation Reviewer | Audits an existing valuation model and proposes sensitivities |
| General Ledger Reconciler | Reconciles the general ledger against source transactions |
| Month End Closer | Drives the monthly close procedure with checklists and journals |
| Statement Auditor | Audits financial statements against IFRS / US GAAP standards |
The targeting is precise: these are the most time-consuming tasks in the middle/back office (pitch, KYC, close) plus a few front-office tasks (meeting prep, earnings). Anthropic is targeting the 6-12 million white-collar finance workers worldwide, 60% of whom still spend more than 30% of their time on these repetitive workflows.
Microsoft 365 integration: Claude becomes a cross-Office agent
This is probably the most structurally important announcement. Claude is no longer a chat with a copy-pasted file: it becomes a cross-application agent that shares context across Excel, PowerPoint, Word, and Outlook.
- Excel — Builds financial models from filings, audits formulas across linked workbooks, runs automated sensitivity analyses
- PowerPoint — Drafts decks that update automatically when underlying numbers change
- Word — Edits credit memos against a firm's internal templates
- Outlook — Beta: reads inbox threads, summarizes, generates contextual replies
The key piece is the shared context: an analyst can start an analysis in Excel, ask Claude to generate the matching PowerPoint deck, then ask Claude to draft the client transmittal email — without ever losing reasoning continuity. Same pattern we identified in our coverage of the Cowork rollout at NEC Japan (30,000 employees on Claude Code) — extended this time to classical Office apps.
Excel, PowerPoint, and Word add-ins are GA today. Outlook is in beta with a GA roadmap targeted for Q3 2026.
The masterstroke: Moody's as a Claude "native app"
The most striking data partnership isn't in the new providers list. It's Moody's.
Moody's embeds its full platform directly into Claude as a native app. In practice: a user can analyze credit ratings and risk data on more than 600 million companies without leaving Claude, without additional authentication, without screen switching.
This joins the existing roster: LSEG, S&P Capital IQ, Morningstar, PitchBook. And adds eight new partners: Verisk, Third Bridge, Fiscal AI, Dun & Bradstreet, Experian, GLG, Guidepoint, IBISWorld.
| Data type | Provider(s) |
|---|---|
| Credit ratings | Moody's (native app), S&P |
| Market data | LSEG, S&P Capital IQ, Morningstar |
| Private markets | PitchBook |
| Expert calls | Third Bridge, GLG, Guidepoint |
| Company data | Dun & Bradstreet, Experian, IBISWorld |
| Insurance risk | Verisk, Fiscal AI |
The result: an analyst preparing a credit memo no longer has to juggle 6 tools + Excel. Everything happens in a single Claude conversation. It mirrors the Bloomberg Terminal model — but with an agent that executes in addition to providing data.
Why this is an existential threat to Bloomberg, FactSet and Capital IQ
Legacy financial terminals run on two fundamentals:
- Data aggregation — You pay to have everything in one place
- Network effect — Every analyst uses it, so you have to use it
This package attacks both.
First, aggregation: Moody's + LSEG + S&P + PitchBook + Morningstar inside Claude is most of the Bloomberg Terminal dataset. Then agency: Claude executes tasks (build me a DCF, audit my formula), not just queries data. Bloomberg Terminal's LP version does have an AI copilot, but it remains constrained by Bloomberg architecture, whose stack was not designed for LLMs.
Pricing is the other angle. Bloomberg Terminal costs ~$30,000/year/user. The equivalent Claude package (with native data app access) reportedly runs at $15,000-$20,000/year/user based on estimates circulating among test firms. A 30-50% data stack cost reduction, with an agentic layer on top.
Bloomberg has not publicly commented yet, but the posture is defensive: internal press notes Mike Bloomberg convened an emergency committee on AI strategy starting the week of May 5.
The timing: why now
Three factors converge to make this move feasible today rather than six months ago.
1. Anthropic's Wall Street revenue trajectory
Per Fortune, JPMorgan, Goldman Sachs, Morgan Stanley and Citi have been Claude enterprise customers since Q4 2025. The ROI measured on pilot tests — 30-40% productivity uplift on pitch and KYC — unlocked significant budgets in early 2026. That revenue base (several hundred million ARR on the finance vertical alone) funds the agent and integration build-out.
2. Available compute
The $200 billion Google Cloud commitment we broke down in our analysis on Anthropic $200B Google Cloud TPU secures inference capacity for high-volume workloads like very large KYC screenings.
3. The OpenAI window
OpenAI announced the same day its partnership with PwC on finance workflows (forecasting, planning, reporting, treasury). Anthropic had to ship before or simultaneously to avoid being boxed into challenger optics. Simultaneous launches create a nascent duopoly on AI financial services.
Competitors and what they have left
| Player | Position May 6, 2026 |
|---|---|
| OpenAI + PwC | Finance services partnership, focus on forecasting and reporting |
| Bloomberg | Legacy terminal + BloombergGPT — but pre-LLM architecture |
| FactSet | AI copilot in beta, 12 months behind the Claude roadmap |
| S&P Capital IQ | Native app inside Claude (turned competitor partner) |
| Refinitiv (LSEG) | Native app inside Claude — strategic pivot |
| Microsoft Copilot Finance | Existing — direct competitor on Excel, but lacks the native Moody's dataset |
Microsoft is the most paradoxical case: Anthropic deeply integrates Office 365 while remaining a competitor to Microsoft Copilot. That's a co-opetition pattern we've seen before with Apple-Google on search. Microsoft accepts because Office 365 + Claude usage friction has gotten too high for enterprise customers. Refusing the integration would cost more than accepting it.
What this changes for SaaS vendors and vertical AI startups
The announcement has two immediate consequences for startups and SaaS vendors:
1. Pure-play vertical finance SaaS is under pressure
If Claude natively offers DCF building, KYC, and month-end close, startups like Rogo or Hebbia must reposition — either toward proprietary, deeply vertical data (like Rogo and its Felix finance agent on a $160M Series D), or toward highly customized enterprise deployment (compliance, legacy systems, data residency).
2. The distribution layer becomes the real moat
For developers building AI tools, the message is clear: the model layer is now commoditized. The moat shifts to distribution + monetization. This is exactly the thesis behind native SDKs like @idlen/chat-sdk, which lets AI apps monetize conversations through contextual ads. For a complete guide to this logic, see our guide to monetizing an AI app and our API economy in 2026 analysis.
Risk areas
Three open issues remain:
- Regulatory compliance. KYC and accounting audit workflows touch strict regulations (BSA/AML, SOX, IFRS). Anthropic ships guard-rails (always human-in-the-loop on close, full audit trail), but no regulator has certified usage yet. A KYC false positive can cost a bank tens of millions in penalties.
- Data residency. European and UK firms must guarantee that customer data does not leave the EU. Anthropic offers EU zones on AWS and GCP, but the Moody's partnership requires individual review.
- Legacy terminal pushback. Bloomberg, FactSet, and Capital IQ will retaliate — probably via pricing and AI exclusivity partnerships (rumor of a Bloomberg-OpenAI deal in late Q2 2026). The war just started.
Conclusion: Anthropic becomes a vertical software vendor
This move confirms Anthropic's mutation. The lab is no longer just selling a model or an API. It is selling vertical packaged business solutions with data integrations and executable templates. That's exactly the Salesforce Industries or SAP Industries logic — applied to generative AI.
For founders wondering what to build in the agent era, the lesson is sharp: generic verticals are now claimed by frontier labs. To break through, you need either proprietary datasets, alternative distribution, or a revenue model that does not depend on the LLM API. The three patterns are compatible, which is why we see in parallel the rise of Sierra at $950M on customer-service agents and Rogo at $1.5B on the Felix finance agent — each with their own dataset, workflow, and distribution.
To watch in the coming months: Bloomberg's response, expansion of the program to other verticals (health, legal), and Anthropic's potential Investor Day around the October IPO we dissect in our $900B valuation analysis.
For more, also read our Claude Opus 4.7 at 87% on SWE-Bench analysis, our Cursor at $50B coverage, and the guide to integrating an AI API in your project.
