Key Takeaways
- Claude now handles full finance workflows across Excel and PowerPoint in a single continuous session, launched February 24, 2026
- Five Anthropic-built plugins cover financial analysis, investment banking, equity research, private equity, and wealth management
- FactSet and MSCI MCP connectors deliver real-time institutional-grade market data directly into Claude’s context
- Wall Street Prep’s 2026 benchmark ranked Claude second overall with a score of 5.5 out of 10, ahead of Copilot (4.4) and ChatGPT (2.5)
Finance workflows have never broken cleanly at the tool boundary, but AI has mostly ignored that reality until now. On February 24, 2026, Anthropic pushed Claude Cowork into the core of how finance actually operates: research, modeling, and presentation in a single uninterrupted session. This is not a chatbot bolted onto a spreadsheet. It is context-carrying, cross-app orchestration that reduces what used to take analysts hours to a fraction of that time.
From Data to Deliverable Without Switching Tools
The most structurally significant update is Claude’s ability to move between Excel and PowerPoint while retaining full context. An equity analyst can ask Claude to analyze earnings, update a financial model in Excel, then build a summary slide in PowerPoint, all without switching tools or losing context along the way.
When inputs change, Claude automatically updates the rest of the workflow. This cross-app continuity is available in research preview for all paid plans on Mac and Windows.
The 5 Finance Plugins and What Each One Does
All five Anthropic-built plugins are available now in the public finance plugin repository and can be installed directly or customized for firm-specific workflows, terminology, and needs.
- Financial analysis: Supports baseline workflows all finance analysts need, covering market and competitive research, financial modeling, and PowerPoint template creation and quality checking
- Investment banking: Helps bankers with deal workflows, including reviewing transaction documents, building comparable company analyses, and preparing pitch materials
- Equity research: Accelerates research workflows by parsing earnings transcripts, updating financial models with new guidance, and drafting research notes
- Private equity: Built for deal sourcing and diligence; Claude reviews large document sets, extracts standardized financial data, models scenarios, and scores opportunities against investment criteria
- Wealth management: Helps advisors analyze portfolios, identify drift and tax exposure, and generate rebalancing recommendations at scale
Institutional Data Directly Into Claude’s Context
Two new MCP connectors bring real data infrastructure into Claude. The FactSet connector provides real-time market data, fundamental analysis, earnings estimates, and research insights to enhance financial analysis and investment research workflows. The MSCI connector brings MSCI’s proprietary index data into Claude, enabling users to access index performance, exposures, constituents, and methodologies.
Partner-built plugins extend this further. S&P Global’s Capital IQ Pro plugin enables company tear sheets, industry transaction summaries, and earnings call previews. The LSEG plugin gives financial professionals access to live yield curves for DCF models, real-time news for morning notes, current cross-asset pricing for portfolio rebalancing, and actual financing economics for deal analysis, and is available for users with active LSEG data entitlements.
How Claude Performs Against Competing AI Finance Tools
Wall Street Prep tested Claude (Opus 4.6), Shortcut (v7.4), Microsoft Copilot (GPT-5 in Agent Mode), and ChatGPT (5.2) by asking each to build a fully integrated three-statement model for Apple using real SEC filings and consensus forecasts. The task is a standard investment banking assignment that a strong analyst typically completes in 2 to 3 hours.
| Subcategory | Winner | Shortcut | Claude | Copilot | ChatGPT |
|---|---|---|---|---|---|
| Preparation and Speed | Shortcut | 10 | 10 | 7 | 0 |
| Formatting | Shortcut | 7 | 5 | 1 | Not scored |
| Accuracy of Historical Data, Assumptions and Sourcing | Copilot | 5 | 3 | 8 | Not scored |
| Sourcing and Commenting | Claude | 5 | 7 | 0 | Not scored |
| Flowthrough of Assumptions | Copilot | 7 | 6 | 8 | Not scored |
| Income Statement Forecasting | Shortcut / Claude | 8 | 8 | 5 | Not scored |
| Circularity | Tie (all failed) | 0 | 0 | 0 | 0 |
| Balance Sheet Forecasting and Integration | Copilot | 5 | 5 | 6 | Not scored |
| Overall Score | Shortcut | 5.9 | 5.5 | 4.4 | 2.5 |
Claude and Shortcut both asked thoughtful clarifying questions after receiving the prompt, closely resembling what WSP expects from a good junior analyst. Copilot and ChatGPT asked none. Claude was the only tool to backsolve EBITDA correctly and provided the best explanations of where data came from and why certain modeling decisions were made.
A significant finding across all tools was data accuracy. Both Claude and Shortcut hallucinated portions of historical data, with errors subtle enough to be dangerous as slightly incorrect line items that still summed to correct subtotals. WSP noted that fixing this would require cell-by-cell auditing that takes longer than inputting the numbers manually.
Considerations Before Deploying in Production
Claude Cowork’s cross-app Excel and PowerPoint integration launched in research preview, meaning it is not at full production readiness. The LSEG plugin requires active data entitlements with LSEG, and the same applies for MSCI and FactSet connectors.
WSP’s testing confirmed that even the best-performing tool, Shortcut, underperformed compared to a lower-tier analyst on the same scoring scale. All four tools scored zero on circularity handling. WSP explicitly cautioned that these tools are strong at kickstarting a project but are not reliable end-to-end companions, and that trusting them too deeply can quietly create problems that take longer to fix than doing the job without AI.
Getting Started With Claude Cowork Finance Plugins
Setup is straightforward for Claude paid plan users.
- Visit the public finance plugin repository to install any of the five Anthropic-built plugins directly or customize them for your firm
- Visit the connector directory in Claude’s settings to authorize FactSet, MSCI, or other enterprise data connections; admins can bundle these into plugins for their teams
- Download the Claude add-in for Excel and the Claude add-in for PowerPoint to enable cross-app workflows
- Partner plugins from LSEG and S&P Global are available in the same financial services plugins repository
- For Cowork and plugin updates across all business functions beyond finance, see Anthropic’s companion blog post
Frequently Asked Questions (FAQs)
What is Claude Cowork and how does it relate to finance?
Claude Cowork is Anthropic’s enterprise AI collaboration platform. The February 24, 2026 update introduced five finance-specific plugins, FactSet and MSCI MCP connectors, and cross-app workflows that carry context from Excel to PowerPoint without losing session state during multi-step finance tasks.
Which finance plugins did Anthropic release for Claude Cowork in 2026?
Anthropic released five plugins: financial analysis, investment banking, equity research, private equity, and wealth management. All five are available in the public finance plugin repository and can be installed directly or customized for a firm’s specific workflows, terminology, and needs.
How does Claude work across Excel and PowerPoint?
Claude carries full context between both apps in a single continuous session. An analyst can instruct Claude to analyze earnings, update a financial model in Excel, and then build a summary slide in PowerPoint without switching tools. When inputs change, Claude automatically updates the rest of the workflow. This feature is in research preview for all paid plans.
What data connectors does Claude Cowork support for finance?
Claude supports MCP connectors for FactSet (real-time market data, fundamentals, earnings estimates, research insights) and MSCI (proprietary index data, performance, exposures, constituents, and methodologies). Partner plugins from S&P Global (Capital IQ Pro) and LSEG (live yield curves, cross-asset pricing, real-time news) are also available.
How does Claude compare to Microsoft Copilot for financial modeling?
In Wall Street Prep’s 2026 benchmark, Claude scored 5.5 overall versus Copilot’s 4.4 out of 10. Claude outperformed Copilot in speed and intent understanding, income statement forecasting, and sourcing and commenting. Copilot outperformed Claude in historical data accuracy and flowthrough of assumptions.
Does Claude hallucinate data in financial models?
Yes, and this is a documented risk. WSP’s testing found that Claude hallucinated portions of historical financial data, with errors subtle enough that they added up to correct subtotals while individual line items were wrong. WSP recommends analysts upload clean PDFs and spreadsheets rather than relying on agents to independently retrieve data.
Does Claude replace finance analysts with these new plugins?
Not currently. WSP’s 2026 benchmark confirmed that even the best AI tool tested underperforms a lower-tier analyst on their standard scoring scale. All tools failed circularity handling completely. WSP describes these tools as strong for kickstarting projects but not as reliable end-to-end companions without extensive human review.
What are the main limitations of Claude Cowork for finance?
The cross-app Excel and PowerPoint feature is in research preview and not fully production-ready. LSEG, MSCI, and FactSet integrations require existing data entitlements. Claude hallucinated historical financial data in WSP testing. All AI tools tested, including Claude, scored zero on circularity handling and none modeled shares outstanding correctly.
