← Back to Claude Mastery
Claude Mastery · Chapter 10 of 30
10

Document Intelligence

Master PDF, Word, Excel, and More

⏱️ 12 min read ◆ 📊 Intermediate ◆ 🎯 Document Analysis

You have a 60-page market research report. Your stakeholders want a 5-point summary by end of day. In the past, that task took hours. With Claude's document intelligence, it takes minutes.

Quick Answer

Claude Document Intelligence means uploading PDFs, Word documents, Excel spreadsheets, CSVs, and more — then asking Claude to read, extract, analyze, compare, and transform the content. Claude genuinely understands document structure, not just raw text, so it can answer specific questions, pull out data points, compare multiple files, and reformat content for different audiences. The key is using a layered approach: start with an overview, then extract specifics, then apply findings to your situation.

Claude can read, analyze, extract, compare, and transform documents. Not just skimming text, but understanding structure, context, and meaning — turning what used to be hours of manual work into a focused conversation.

What Can Claude Actually Do with Uploaded Documents?

✅ Claude Does This Well
  • Reading & Comprehension — Document structure, main argument, key themes
  • Extraction — Specific data points, tables, named entities, figures
  • Analysis — Compare documents, spot patterns, evaluate arguments
  • Transformation — Summarize, reformat, convert tone or structure
  • Synthesis — Combine information from multiple sources
❌ Know the Limitations
  • Scanned PDFs without OCR — Images of text may not be readable
  • Very large files — Extremely long documents may be truncated
  • Live Excel formulas — Reads values, not formula logic
  • Password-protected files — Cannot open locked documents
  • Real-time data — Uploaded documents are static, not live

How Do You Get the Most from Claude When Analyzing PDFs?

PDFs are the most common document type uploaded to Claude. Four strategies work best:

Strategy 1
The Layered Question Approach

Don't ask everything at once. Start broad, then go deep across three rounds:

Round 1 — Understand the document:

I've uploaded a market research report called "Global SaaS Trends 2024". Before I ask specific questions: 1. What is the main argument or finding of this report? 2. How is it structured? (List the main sections) 3. What time period does it cover? 4. Who is the intended audience? 5. What are the top 3 things I should know from this document?

Round 2 — Extract specific information:

Based on this report, please answer: 1. What growth rate does it project for the CRM segment specifically? 2. Which geographic market shows the fastest growth? 3. What does it say about AI integration in SaaS products?

Round 3 — Apply to your situation:

Our company is a B2B SaaS in the HR tech space, Series A stage. Based on what you've read in this report, what are the top 3 strategic implications for us specifically?

Each round builds on the last. Claude's understanding deepens with each layer of questioning.

Strategy 2
The Targeted Extraction Template

When you need specific data from a long document:

From the uploaded [document name], please extract: SECTION: [Specific section or chapter] Find and list: 1. All statistics and data points (with page/section reference) 2. All recommendations or action items 3. All risk factors mentioned 4. Named companies or case studies used as examples Format as a structured table where possible.
Strategy 3
Multi-PDF Comparison

Upload multiple documents and compare them side by side:

I've uploaded three vendor proposals: - proposal-vendor-a.pdf - proposal-vendor-b.pdf - proposal-vendor-c.pdf Please compare them across: 1. Pricing structure (monthly, annual, per-seat) 2. Implementation timeline 3. Support and SLA commitments 4. Key differentiators each vendor claims 5. Risks or concerns in each proposal Create a comparison table, then give me your assessment of which vendor appears strongest and why.
Strategy 4
Document Q&A

Treat your PDF like a searchable knowledge base — especially powerful for contracts and legal documents:

I'm going to ask you several questions about the uploaded contract. Please answer each one with the exact clause or section that supports your answer. Questions: 1. What is the payment terms clause? 2. What are the termination conditions? 3. Is there an exclusivity clause? If so, what does it say? 4. What jurisdiction governs this contract? 5. What are the liability limitations?

How Do You Work with Word Documents in Claude?

Word documents often contain structured content — reports, proposals, meeting notes, drafts. Three use cases stand out:

Use Case 1: Document Review and Feedback

I've uploaded a business proposal draft called "proposal-draft-v2.docx". Please review it and provide feedback on: STRUCTURE: - Does the narrative flow logically? - Are sections in the right order? - Is anything missing? CONTENT: - Is the value proposition clear? - Are the claims substantiated? - Any gaps in the argument? TONE & STYLE: - Is it appropriate for an enterprise audience? - Any sections that feel too informal or too jargon-heavy? QUICK WINS: - Top 3 changes that would most improve this proposal Be specific — reference the actual sections when giving feedback.

Use Case 2: Document Transformation

Turn one format into multiple versions for different audiences:

I've uploaded our annual report (annual-report-2024.docx). Transform this into: 1. A 3-paragraph executive summary (for the board) 2. A 10-bullet employee update (for all-hands meeting) 3. A 5-tweet thread (for social media announcement) 4. A 2-minute talking points script (for CEO presentation) Each version should capture the key message but be optimized for its specific audience and format.

Use Case 3: Meeting Notes to Action Items

I've uploaded notes from our Q3 planning meeting (meeting-notes-q3.docx). From these notes, extract and organize: DECISIONS MADE: [All confirmed decisions with owner if mentioned] ACTION ITEMS: [Table: Task | Owner | Deadline | Priority] OPEN QUESTIONS: [Items discussed but not resolved] FOLLOW-UP MEETINGS NEEDED: [Topics that need a separate discussion] KEY CONTEXT FOR THOSE WHO WEREN'T THERE: [3-5 bullet summary of most important outcomes]

How Do You Analyze Excel and Spreadsheet Data with Claude?

Excel and CSV files unlock some of Claude's most powerful analytical capabilities. When you upload a spreadsheet, Claude can read all values across sheets and tabs, understand column headers and data structure, perform analysis, identify patterns and anomalies, create narrative summaries from numbers, and generate code to process data further.

Key Limitation

Claude reads spreadsheet values, not live formula logic. It cannot update files in real-time or connect to live data sources. For complex financial models, verify key outputs against the source.

Strategy 1: Data Overview First

Always start by getting Claude's read on the data structure before any analysis:

I've uploaded our sales data spreadsheet (sales-2024.xlsx). Before any analysis, please give me: 1. What data is in each tab/sheet? 2. What time period does this cover? 3. What are the key column headers and what do they represent? 4. Any obvious data quality issues (missing values, inconsistencies)? 5. What analysis would be most valuable based on what you see?

Strategy 2: Structured Analysis Request

Using the uploaded sales-2024.xlsx, please analyze: PERFORMANCE OVERVIEW: - Total revenue by quarter - Month-over-month growth rate - Best and worst performing months PRODUCT ANALYSIS: - Revenue breakdown by product category - Top 5 products by revenue and by unit volume REGIONAL BREAKDOWN: - Revenue by region - Which region is growing fastest - Any regions showing decline ANOMALIES: - Unusual spikes or drops - Months that deviate significantly from trend Format with clear headers. Include specific numbers for each finding.

Strategy 3: Excel to Written Narrative

Turn numbers into a boardroom-ready report:

I've uploaded our monthly KPI dashboard (kpi-dashboard.xlsx). Write an executive summary (400 words max) that: - Opens with the headline story (what's the most important thing?) - Covers the 3 most significant metrics and what they mean - Explains any concerning trends and potential causes - Ends with 2-3 recommended actions based on the data Write it for our board of directors — assume they understand the business but haven't seen this month's numbers.

Strategy 4: Data Comparison Across Periods

I've uploaded two files: - q2-results.xlsx (April-June 2024) - q3-results.xlsx (July-September 2024) Compare these two quarters across revenue (total change, product and regional breakdown), efficiency (cost of sales, gross margin), and customer metrics (new acquisition, churn, ACV). What's the most significant change between Q2 and Q3? What's driving it?

What Are the Most Powerful Multi-Document Workflows?

Some of the most valuable document intelligence work involves multiple documents together. Three workflows stand out:

Workflow 1: Research Synthesis

I've uploaded 5 research papers on remote work productivity. Please synthesize these into: CONSENSUS FINDINGS: What do all (or most) papers agree on? CONTRADICTIONS: Where do papers disagree? What might explain it? STRONGEST EVIDENCE: Which findings have the most robust support? PRACTICAL IMPLICATIONS: Based on this research, what are 5 actionable recommendations for a company implementing remote work? GAPS: What questions remain unanswered by this body of research?

Workflow 2: Competitive Intelligence

I've uploaded annual reports from 3 competitors. Analyze their strategic priorities: 1. What growth areas is each company investing in? 2. What risks does each company acknowledge? 3. How does each company describe their competitive advantage? 4. Revenue and growth comparison (if data available) 5. Which company appears to be executing most effectively and why?

Workflow 3: Contract Portfolio Review

I've uploaded 5 client contracts. Please review them collectively: CONSISTENCY CHECK: - Which payment terms vary across contracts? - Which contracts have non-standard clauses? - Where are the liability limits different? RISK ASSESSMENT: - Which contracts have the most favorable terms for us? - Which have unusual risks or obligations? STANDARDIZATION OPPORTUNITIES: - What would a standard template look like based on these?

What Are the Best Practices for Claude Document Analysis?

1. Name Your Documents Clearly

❌ "I uploaded the file" ✅ "I uploaded our Q3 sales report (q3-sales-report.xlsx)"

Claude references documents by name — clear naming avoids confusion when multiple files are in play.

2. State Your Purpose and Role Upfront

❌ "Analyze this document" ✅ "I'm the CFO preparing for a board Q&A. Analyze this financial report and identify the 3 assumptions board members are most likely to challenge and how I should respond to each."

The same document looks different depending on who's reading it. Your role shapes what analysis is most relevant.

3. Ask for Citations

✅ "When answering, please reference the specific section or page where you found each piece of information."

Especially important for contracts, research, and compliance documents.

4. Verify Critical Numbers

✅ "Double-check these key numbers against the source document: - Total Q3 revenue: $4.2M - YoY growth: 23% - Top product: Widget Pro ($1.8M)"

Claude is accurate but not infallible. For critical financial figures, always verify against the source.

5. Iterate with Follow-Up Questions

First: "What are the main findings?" Then: "Tell me more about finding #2" Then: "What evidence supports that finding?" Then: "How does this compare to what we saw in the other report?"

Documents reward iterative exploration. Each follow-up question adds depth and precision.

What Are the Most Common Document Analysis Mistakes to Avoid?

Mistake 1: Vague Opening Question

❌ "What do you think of this document?" ✅ "I've uploaded our 5-year strategic plan. I'm the CFO preparing for a board Q&A. Identify the 3 financial assumptions board members are most likely to challenge and suggest how I should respond."

Mistake 2: No Context About Yourself

✅ "I'm a first-time founder reviewing this term sheet. Focus on terms that are unusual or potentially unfavorable for founders." vs. ✅ "I'm a VC associate reviewing this term sheet. Focus on any investor-friendly protections that might be missing."

Mistake 3: Treating Claude's Output as Final

Always review Claude's document analysis, especially for legal documents (have a lawyer review), financial figures (verify against source), medical information (consult professionals), and regulatory compliance (expert review required).

Mistake 4: One Giant Question

❌ One enormous prompt asking 15 different things simultaneously ✅ Layered questions, each building on previous answers
Key Takeaways
  • Documents unlock deep analysis — Not just text reading, but genuine comprehension of structure and meaning
  • Start with document overview — Understand structure before diving into specifics
  • Layer your questions — Broad first, then specific, then applied to your situation
  • Multiple documents compound value — Comparison and synthesis are where the real power lies
  • Always provide your context — Who you are shapes what analysis is most relevant
  • Cite and verify — Ask for references; double-check critical figures
  • Right strategy per type — PDFs, Word, and Excel each have optimal approaches
Your Turn: Assignment

Challenge: Work through a real document you have.

Beginner: Upload a PDF report you've been meaning to read. Use the layered question approach to extract the 5 most relevant insights for your work.

Intermediate: Upload an Excel or CSV file from your work. Ask Claude to perform the structured analysis — overview → performance → anomalies → narrative.

Advanced: Upload 2–3 related documents (multiple reports, competitor analyses, or research papers). Use the multi-document synthesis workflow to produce a comparative analysis you couldn't have done as quickly on your own.

Share: What was the most surprising insight Claude found in your document?

Frequently Asked Questions

Claude can read and analyze PDFs, Word documents (.docx), Excel spreadsheets (.xlsx), CSV files, PowerPoint presentations, and plain text files. It can understand document structure, extract specific data points, compare multiple documents, and transform content from one format to another. The main limitations are scanned PDFs without OCR, password-protected files, and very large documents that may be truncated.

Always start with a document overview before asking specific questions. Ask Claude to describe the document structure, main argument, key sections, and what analysis would be most valuable. This prevents wrong assumptions and aligns Claude with your actual data. The layered question approach works best: broad understanding first, then specific extraction, then applied analysis for your situation.

Yes. You can upload multiple documents in the same conversation and ask Claude to compare them across specific dimensions. This works well for comparing vendor proposals, analyzing competitor annual reports, synthesizing multiple research papers, and reviewing contract portfolios. Specify the exact comparison criteria you need rather than asking for a general comparison.

When you upload an Excel file or CSV, Claude can read all values across sheets and tabs, understand column headers and data structure, perform analysis on the data, identify patterns and anomalies, create narrative summaries from numbers, and generate code to process the data further. It cannot execute live Excel formulas — it reads the values but not formula logic — and it cannot connect to live data sources.

Name your documents clearly when referencing them. State your purpose and role upfront so the analysis is relevant to your situation. Ask for citations so Claude references the specific sections supporting each answer. Verify critical numbers against the source document. Use layered follow-up questions rather than asking everything in one giant prompt. For legal, financial, and medical documents, always have a qualified professional review Claude's analysis.

View All Chapters