AI Document Analysis

Automatically classify and analyze documents with AI-powered intelligence.

What is AI Document Analysis?

AI Document Analysis uses artificial intelligence to understand your documents instantly:

Auto-Classification: Categorize documents (contract, invoice, legal, personal, etc.)
Smart Summaries: Generate human-readable summaries in seconds
Entity Extraction: Identify names, dates, amounts, and organizations
Key Points: Extract the most important information
Confidence Scoring: Know how confident the AI is about its analysis

Why Use It?

Save Time:

Analyze 100 documents in minutes (not hours)
No manual reading required for initial classification
Instant summaries for quick review

Improve Organization:

Auto-tag documents by type
Extract metadata for search
Group related documents automatically

Find Information Faster:

Search by extracted entities
Filter by document category
Locate key information in seconds

Document Classification

How It Works

Upload a document and AI automatically classifies it:

Supported Categories:

Contract: Employment contracts, service agreements, NDAs
Invoice: Bills, receipts, payment records
Legal: Court documents, legal notices, filings
Report: Business reports, research papers, analyses
Correspondence: Letters, memos, emails
Financial: Bank statements, tax documents, financial reports
Personal: Personal letters, diaries, notes
Other: Uncategorizable documents

Example Analysis:

{
  "category": "contract",
  "confidence": 0.95,
  "summary": "Employment agreement between TechCorp and Jane Smith for Senior Software Engineer position, starting January 15, 2024.",
  "keyPoints": [
    "Position: Senior Software Engineer",
    "Salary: $150,000 annually",
    "Start Date: January 15, 2024",
    "Benefits: Health insurance, 401(k), stock options",
    "At-will employment"
  ],
  "entities": {
    "names": ["TechCorp", "Jane Smith"],
    "dates": ["2024-01-15"],
    "amounts": ["$150,000"]
  }
}

Using Classification

Automatic Classification:

Enable Settings → AI Preferences → Auto-Classification
Upload document to vault
AI classifies automatically in background
View results in document details

Manual Classification:

Go to Documents → Select document
Click Analyze with AI button
Review classification results
Accept or modify classification

Bulk Classification:

Go to Documents → Select multiple documents
Click Bulk Actions → Classify with AI
AI processes all selected documents
Review results summary

Confidence Scores

Understanding Confidence:

0.95-1.00: Extremely confident (very likely correct)
0.85-0.94: High confidence (usually correct)
0.70-0.84: Medium confidence (review recommended)
Below 0.70: Low confidence (manual review required)

What Affects Confidence:

Document clarity and quality
Standard vs. custom document formats
Language complexity
Document length and structure

Low Confidence? Tips:

Upload higher quality scans
Use OCR-processed PDFs (searchable text)
Provide context (add document title/description first)
Split combined documents (multiple contracts in one file)

Deep Document Analysis

Advanced Analysis Features

Beyond Classification:

Detailed content analysis
Relationship mapping between entities
Chronological event extraction
Risk and compliance flagging
Similar document detection

Use Cases:

Legal Review: Identify key clauses, obligations, deadlines
Contract Management: Extract terms, renewal dates, parties
Financial Analysis: Extract transactions, amounts, account numbers
Research: Summarize papers, extract key findings, citations

Using Deep Analysis

Analyze a Document:

Go to Documents → Select document
Click AI Analysis → Deep Analysis
Wait for processing (10-30 seconds)
Review detailed analysis results

Analysis Output:

{
  "summary": {
    "brief": "One-paragraph overview",
    "detailed": "Multi-paragraph analysis with context"
  },
  "entities": {
    "people": ["Jane Smith (Employee)", "John Doe (CEO)"],
    "organizations": ["TechCorp Inc.", "Acme Legal Services"],
    "locations": ["San Francisco, CA", "Remote"],
    "dates": [
      {"date": "2024-01-15", "context": "Employment start date"},
      {"date": "2025-01-15", "context": "First review date"}
    ],
    "amounts": [
      {"amount": "$150,000", "context": "Annual salary"},
      {"amount": "$5,000", "context": "Signing bonus"}
    ]
  },
  "keyTerms": {
    "obligations": [
      "Confidentiality agreement",
      "Non-compete clause (12 months)",
      "Intellectual property assignment"
    ],
    "rights": [
      "Health insurance coverage",
      "Stock options (10,000 shares, 4-year vesting)",
      "Unlimited PTO policy"
    ],
    "deadlines": [
      {"date": "2024-01-10", "task": "Complete background check"},
      {"date": "2025-01-15", "task": "Performance review"}
    ]
  },
  "riskFactors": [
    "Non-compete clause may be unenforceable in California",
    "No specified severance terms"
  ],
  "relatedDocuments": [
    "Employment Offer Letter - Jane Smith.pdf",
    "TechCorp Employee Handbook 2024.pdf"
  ]
}

Entity Extraction

What Are Entities?

Entities are important pieces of information extracted from documents:

Supported Entity Types:

Names: People, organizations, products
Dates: Specific dates, date ranges, deadlines
Amounts: Money, percentages, quantities
Locations: Addresses, cities, countries, regions
Contact Information: Emails, phone numbers, websites
Identifiers: Account numbers, case numbers, reference IDs

Using Extracted Entities

Search by Entity:

Search: "Jane Smith"
Results: All documents mentioning Jane Smith

Search: "amount:>$100,000"
Results: Documents with amounts over $100,000

Search: "date:2024-01"
Results: Documents with dates in January 2024

Entity Timeline:

View all extracted dates chronologically
See which documents contain specific dates
Create timelines across multiple documents

Entity Relationships:

Map connections between people and organizations
Track document flow between parties
Visualize entity networks

Key Points Extraction

What Are Key Points?

Key Points are the most important takeaways from a document:

Examples by Document Type:

Contract:

Parties involved
Key obligations and rights
Payment terms and amounts
Important dates and deadlines
Termination conditions

Invoice:

Invoice number and date
Billed items and quantities
Total amount and payment terms
Billing party and recipient
Due date

Legal Document:

Case or matter number
Parties involved
Key claims or defenses
Important rulings or decisions
Next action dates

Report:

Main findings or conclusions
Key statistics and data points
Recommendations
Executive summary points
Methodology highlights

Using Key Points

Quick Review:

Open document in viewer
View Key Points panel (right sidebar)
Review 3-10 bullet points
Click any point to jump to source in document

Export Key Points:

Markdown: For notes and wikis
PDF: Professional summary document
CSV: Spreadsheet-compatible format
JSON: Programmatic access

Share Key Points:

Share summary without sharing full document
Create case briefing from key points
Generate reports from extracted points

Document Summaries

Summary Types

Brief Summary (2-3 sentences):

High-level overview
Main purpose or topic
Primary parties or entities

Example:

"Employment agreement between TechCorp and Jane Smith for Senior Software Engineer role. Outlines salary ($150K), benefits, and standard employment terms. Effective January 15, 2024."

Detailed Summary (1-2 paragraphs):

Comprehensive overview
Context and background
Key details and terms
Important conditions or clauses

Example:

"This employment agreement establishes the terms of employment between TechCorp Inc. and Jane Smith for the position of Senior Software Engineer, effective January 15, 2024. The agreement specifies an annual salary of $150,000, comprehensive benefits including health insurance and stock options (10,000 shares with 4-year vesting), and unlimited PTO. The position is at-will employment, with standard confidentiality, non-compete (12 months), and intellectual property assignment clauses. The agreement includes provisions for performance reviews, professional development stipend ($2,000 annually), and remote work flexibility."

Using Summaries

View Summary:

Document details page → AI Analysis tab
Document list view → Hover for quick summary
Search results → Summary preview

Customize Summaries:

Length: Brief, Standard, Detailed
Focus: General, Legal, Financial, Technical
Audience: Executive, Legal, Technical, General

Export Summaries:

Generate PDF with summary + document
Export summaries for multiple documents
Create executive briefing from summaries

Practical Examples

Example 1: Organizing Legal Documents

Scenario: You uploaded 50 mixed legal documents to your vault.

Steps:

Select all 50 documents
Bulk Actions → Classify with AI
AI categorizes: 20 contracts, 15 invoices, 10 legal filings, 5 correspondence
Auto-tag documents based on classification
Review confidence scores, manually adjust low-confidence classifications
Search and filter by category

Time Saved: 50 documents × 5 minutes each = 250 minutes manual → 10 minutes with AI

Example 2: Contract Analysis

Scenario: Reviewing employment contract before signing.

Steps:

Upload contract PDF
Analyze with AI → Deep Analysis
Review key points:
- Salary and compensation
- Benefits and perks
- Obligations and restrictions
- Termination conditions
Check risk factors highlighted by AI
Extract important dates and deadlines
Export analysis to share with attorney

Insight: AI flags non-compete clause that may be unenforceable in your state.

Example 3: Invoice Processing

Scenario: Process monthly invoices from multiple vendors.

Steps:

Upload 30 invoices
AI auto-classifies as "invoice"
Extract key information:
- Invoice numbers
- Due dates
- Amounts
- Vendor names
Export to CSV
Import to accounting software

Time Saved: 30 invoices × 3 minutes each = 90 minutes manual → 5 minutes with AI

Best Practices

When to Use AI Analysis

✅ Good Use Cases:

High volume of similar documents
Initial document review and categorization
Extracting structured data from unstructured documents
Creating summaries for quick reference
Finding specific information across many documents

❌ Not Recommended For:

Final legal review (always use human attorney)
Highly sensitive decisions based solely on AI
Documents in poor quality or non-standard formats
Languages other than English (accuracy may vary)

Improving Accuracy

Document Quality:

Upload high-resolution scans (300+ DPI)
Use searchable PDFs (OCR processed)
Avoid handwritten documents (low accuracy)
Ensure text is machine-readable

Provide Context:

Add document title and description
Tag documents with relevant metadata
Group related documents in folders
Use consistent naming conventions

Review Results:

Always check confidence scores
Verify extracted entities
Confirm key points accuracy
Report errors to improve AI

Privacy Considerations

Data Processing:

AI analyzes document content (requires reading)
Content sent to AI provider (Google Gemini)
No data retained by AI provider
All processing encrypted end-to-end

Sensitive Documents:

Review terms before analyzing confidential documents
Disable AI for highly sensitive vaults
Use manual classification for top-secret materials
Audit AI operations via Settings → AI Logs

Troubleshooting

Low Confidence Scores

Problem: AI returns low confidence (<0.70) for classification.

Solutions:

Improve scan quality: Re-scan at higher resolution
Use OCR: Convert image PDFs to searchable text PDFs
Provide context: Add title and description manually first
Manual review: Classify manually when AI is uncertain

Incorrect Classification

Problem: AI misclassifies document (e.g., invoice classified as contract).

Solutions:

Review extracted text: Ensure OCR read document correctly
Check document format: Standard formats work better
Report error: Settings → Report AI Issue (helps improve AI)
Manual override: Change classification manually

Missing Entities

Problem: AI doesn't extract expected entities (names, dates, amounts).

Solutions:

Check text readability: Ensure document text is machine-readable
Formatting issues: Entities in tables/images may not extract well
Language support: Entities in non-English text may be missed
Manual addition: Add entities manually via Edit Metadata

Slow Processing

Problem: AI analysis takes longer than expected.

Solutions:

Document size: Large documents (100+ pages) take longer
Queue position: Multiple requests may queue
API limits: Rate limits may slow processing
Check status: Settings → AI Operations to monitor progress

FAQ

How accurate is AI classification?

85-95% accuracy for standard document types (contracts, invoices, legal documents). Accuracy varies based on document quality, format, and language.

Can AI read handwritten documents?

Limited support. AI can read clear handwriting but accuracy is 60-70%. For best results, use typed/printed documents or OCR-processed scans.

Does AI work with images?

Yes, but images must contain readable text. AI extracts text via OCR, then analyzes. Photos of documents work if text is clear and legible.

What file formats are supported?

PDF: Best support (native and scanned)
Word: .docx, .doc
Images: .jpg, .png, .tiff (OCR processed)
Text: .txt, .md
Other: .rtf, .odt (limited)

Can I customize AI categories?

Not yet. AI uses predefined categories. Coming Soon: Custom categories for Enterprise customers.

How do I disable AI for specific vaults?

Settings → Vaults → Select vault → AI Settings → Disable AI Analysis

Documents in this vault will not be analyzed by AI.

Can I re-analyze a document?

Yes. Documents → Select document → AI Analysis → Re-analyze

Useful if AI analysis failed or you uploaded a new version.

Are AI results auditable?

Yes. All AI operations are logged:

Settings → AI Logs shows all AI operations
Export logs for compliance (JSON, CSV)
Enterprise: Integrate logs with SIEM

API Access

Classify Document via API

curl -X POST https://api.torvus security.com/v1/ai/classify \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "documentId": "doc_abc123",
    "vaultId": "vault_xyz789"
  }'

Response:

{
  "category": "contract",
  "confidence": 0.95,
  "summary": "Employment agreement between TechCorp and Jane Smith...",
  "keyPoints": ["Salary: $150K", "Start date: Jan 15, 2024"],
  "entities": {
    "names": ["TechCorp", "Jane Smith"],
    "dates": ["2024-01-15"],
    "amounts": ["$150,000"]
  },
  "metadata": {
    "provider": "gemini",
    "model": "gemini-2.0-flash-exp",
    "processingTime": "1.2s"
  }
}

Getting Help

Support Channels:

AI Issues: ai-support@torvussecurity.com
Feature Requests: feature-requests@torvussecurity.com
Documentation: AI Assistant Overview

Resources:

Last Updated: October 15, 2025

What is AI Document Analysis?​

Why Use It?​

Document Classification​

How It Works​

Using Classification​

Confidence Scores​

Deep Document Analysis​

Advanced Analysis Features​

Using Deep Analysis​

Entity Extraction​

What Are Entities?​

Using Extracted Entities​

Key Points Extraction​

What Are Key Points?​

Using Key Points​

Document Summaries​

Summary Types​

Using Summaries​

Practical Examples​

Example 1: Organizing Legal Documents​

Example 2: Contract Analysis​

Example 3: Invoice Processing​

Best Practices​

When to Use AI Analysis​

Improving Accuracy​

Privacy Considerations​

Troubleshooting​

Low Confidence Scores​

Incorrect Classification​

Missing Entities​

Slow Processing​

FAQ​

How accurate is AI classification?​

Can AI read handwritten documents?​

Does AI work with images?​

What file formats are supported?​

Can I customize AI categories?​

How do I disable AI for specific vaults?​

Can I re-analyze a document?​

Are AI results auditable?​

API Access​

Classify Document via API​

Getting Help​

What is AI Document Analysis?

Why Use It?

Document Classification

How It Works

Using Classification

Confidence Scores

Deep Document Analysis

Advanced Analysis Features

Using Deep Analysis

Entity Extraction

What Are Entities?

Using Extracted Entities

Key Points Extraction

What Are Key Points?

Using Key Points

Document Summaries

Summary Types

Using Summaries

Practical Examples

Example 1: Organizing Legal Documents

Example 2: Contract Analysis

Example 3: Invoice Processing

Best Practices

When to Use AI Analysis

Improving Accuracy

Privacy Considerations

Troubleshooting

Low Confidence Scores

Incorrect Classification

Missing Entities

Slow Processing

FAQ

How accurate is AI classification?

Can AI read handwritten documents?

Does AI work with images?

What file formats are supported?

Can I customize AI categories?

How do I disable AI for specific vaults?

Can I re-analyze a document?

Are AI results auditable?

API Access

Classify Document via API

Getting Help