Skip to main content

​
Overview

Arbiter’s document analysis system provides comprehensive AI-powered review of contracts and legal documents with:
  • Structure extraction and section identification
  • Multi-party perspective analysis
  • Clause-level risk assessment
  • Numerical scoring with configurable algorithms
  • Executive summaries and insights

​
Analysis Workflow

1

Upload Your Document

Upload a contract (PDF, DOCX, or TXT). Arbiter extracts text and identifies the document structure.
2

Structure Analysis

Click “Analyze Document” to start. AI extracts comprehensive metadata:Document Structure
  • Sections and hierarchy
  • Cross-references between sections
Parties
  • All party names and roles
  • Party relationships
Key Dates
  • Effective date
  • Expiration date
  • Term length
  • Auto-renewal provisions
  • Renewal notice periods
Financial Terms
  • Contract value or value range
  • Currency
  • Payment terms and schedules
Legal Framework
  • Governing law
  • Jurisdiction
  • Dispute resolution mechanism
  • Liability caps
Classifications
  • Document type
  • Industry tags
  • Regulatory frameworks
  • Missing critical sections
3

Configure Analysis Options

Before running deeper analysis, configure your preferences:
  • Advanced Reasoning - Toggle for deeper AI analysis (~1.5x tokens)
  • GitLaw Research - Enable to include case law and statutory references
4

Party Analysis

Select a party perspective and run analysis. Each section is evaluated for:
  • Clarity issues
  • Enforceability concerns
  • Risk factors
  • Party-specific advantages/disadvantages
5

Clause Analysis (Optional)

Deep dive into individual clauses with:
  • Specific issues identified
  • Suggested revisions
  • Risk ratings

​
Analysis Modes

​
Standard Analysis

  • Fast, comprehensive review
  • Single AI model
  • Complete scoring and insights
  • ~5 - 10 minutes for typical contracts

​
Advanced Analysis

  • Deeper reasoning and analysis
  • Enhanced AI reasoning capabilities
  • More detailed insights
  • 1.5x token cost
  • ~10 - 20 minutes for typical contracts
Use Advanced Mode for complex contracts, high-stakes agreements, or when you need the most comprehensive analysis possible.

​
Understanding Scores

Arbiter uses a 0 - 100 scoring system across three dimensions:

​
Clarity Score

Measures how clearly obligations, rights, and terms are expressed. Uses a weighted formula across five sub-components. Metrics analyzed:
  • Undefined Terms (25% weight) - Terms used throughout but never defined
  • Ambiguous Language (25% weight) - Vague phrasing interpretable multiple ways
  • Inconsistencies (20% weight) - Internal contradictions or conflicting statements
  • Structural Issues (15% weight) - Problems with document organization, formatting, or cross-references
  • Factual Concerns (15% weight) - Potentially incorrect or unverifiable factual claims
Linguistic errors (grammar, spelling) are also factored into the structural component.

​
Enforceability Score

Assesses legal validity and enforceability of provisions. Evaluates whether clauses would hold up in court. Metrics analyzed:
  • Problematic Provisions (30% weight) - Clauses that may face enforceability challenges (e.g., penalty clauses, unconscionable terms)
  • Jurisdictional Issues (25% weight) - Provisions that may conflict with local law or practice
  • Missing/Incomplete Elements (25% weight) - Essential provisions that are absent or unfinished
  • Balance Assessment (20% weight) - Significant asymmetry in rights and obligations, or one-sided provisions lacking mutuality
An overall imbalance flag triggers a flat penalty; otherwise, individual one-sided provisions are counted.

​
Risk Score

Evaluates business and legal risks using a diminishing returns formula. Higher score = lower risk exposure. Metrics analyzed:
  • Risk Items - Each identified risk has a severity rating (1 - 5); cumulative weighted severity drives the base score
  • Unusual Provisions - Atypical or surprising terms add a penalty (+4% exposure per item)
  • Safeguards - Protective provisions provide mitigation ( - 3% exposure per item, max 20% reduction)
Risk types include: financial, operational, legal, and reputational exposure.

​
Scoring Methodology

Arbiter offers three scoring approaches:Fixed Window (Legacy)
  • Static 1,000-word baseline
  • Simple and predictable
  • May not account for document length variations
Adaptive Window
  • Adjusts baseline to actual document length
  • Fair comparison across different sized contracts
  • Recommended for most use cases
Adaptive + Bayesian (Default)
  • Combines adaptive window with statistical smoothing
  • Prevents score instability in short sections
  • Most accurate for enterprise use
  1. Metrics Extraction - AI extracts 15 structured metric types (undefined terms, risks, safeguards, etc.) with exact text references
  2. Rate Calculation - Issue counts are normalized per effective window (typically 500 words) to fairly compare sections of different lengths
  3. Sub-Score Computation - Each pillar has weighted sub-components that start at 100 and deduct penalty points per normalized issue
  4. Dampening - Small sections (under 200 words) have dampened penalties; large sections (over 2000 words) have capped penalties
  5. Quantization - Raw scores are clipped to 0 - 100 and rounded to nearest 5 for consistency
  6. Overall Score - Equal-weighted average of Clarity, Enforceability, and Risk scores (33.3% each)

​
Party Perspective Analysis

Arbiter analyzes contracts from each party’s perspective:
1

Party Identification

AI automatically extracts all parties from the contract
2

Multi - Perspective Analysis

Each section is analyzed considering what’s favorable/unfavorable for each party
3

Risk Matrix

View comparative scores showing how each section affects different parties

​
Key Features

​
Executive Summary

AI-generated overview covering:
  • Overall assessment
  • Key risks and opportunities
  • Critical provisions
  • Missing clauses
  • Recommendations

​
Metadata Extraction

Automatically identifies:
  • Effective dates and expiration
  • Contract value and payment terms
  • Governing law and jurisdiction
  • Auto-renewal provisions
  • Notice periods

​
Section Summaries

Each section includes:
  • Plain language summary
  • Key obligations and deadlines
  • Restrictions and limitations
  • Protective measures
  • Compliance requirements

​
Clause - Level Analysis

Deep review of individual clauses with:
  • Specific issue identification
  • Risk level assessment (Low, Medium, High)
  • Suggested revisions
  • Enforceability analysis

​
Analysis Tabs

Overview and risk scoring for each document section. Includes:
  • Section-by-section clarity, enforceability, and risk scores
  • Expandable analysis with detailed findings
  • Filter by category (definitions, obligations, termination, etc.)
  • Party perspective selector

​
Advanced Features

​
Redlining

This feature is being redesigned and will change in an upcoming release.
Apply AI-suggested revisions with track changes:
  • Redline Mode - Shows insertions (green) and deletions (red)
  • Replace Mode - Directly applies changes
  • Revert - Undo applied changes
  • Export - Download with track changes preserved

​
Version Control

Git-style document versioning:
  • Automatic saves on significant changes
  • Manual version creation
  • Compare any two versions
  • Restore previous versions
  • View change history with timestamps

​
Document Intelligence

AI-powered insights including:
  • Missing section detection
  • Regulatory framework identification
  • Industry classification
  • Dispute resolution analysis
  • Liability cap extraction

​
Best Practices

Start with Doc Intelligence

Review the executive summary in the Doc Intelligence tab first to understand the contract’s scope and key findings.

Compare Party Perspectives

Use the party perspective selector in Section Analysis to identify provisions with different impacts per party.

Focus on Red Scores

Sections with scores below 60 (orange/red) need immediate attention.

Review Redlining Suggestions

For critical sections, review the Redlining tab for specific clause-level issues and AI-suggested revisions.

Export for Clients

Generate professional PDF reports with your analysis for client presentations.

​
Limitations

Token Consumption - Analysis uses tokens based on document length and complexity. Token costs are displayed before you run analysis so there are no surprises.
Large Documents - Arbiter handles documents of any length. Larger documents will use more tokens and may take longer to process.

​
Troubleshooting

  • Standard analysis: typically 5 - 10 minutes
  • Advanced analysis: typically 10 - 20 minutes
  • Check your internet connection
  • Refresh the page if stuck over 30 minutes
  • Check which normalization method is selected (Settings → Scoring Methodology)
  • Try “Adaptive + Bayesian” for most accurate results
  • Review the scoring trace to see specific issues identified
  • Some sections may be too short for meaningful analysis
  • Check if the document structure was correctly identified
  • Re-analyze with Advanced Mode for better section detection

​
Next Steps

Document Drafter

Create contracts from scratch with AI

AI Research

Research legal questions with multi-model AI