Contract Analysis

How AI Reads a Contract

When you review a contract, you read it linearly — clause by clause, page by page — building a mental model of obligations, risks, and missing protections. AI does something similar but at machine speed. It breaks the document into segments, identifies clause types (indemnity, limitation of liability, termination, governing law), extracts key data points (dates, amounts, party names), and scores each clause for risk.

Think of it like having a very fast, very literal junior who reads every word but needs your judgment to decide what matters. This chapter shows you how that process works, what it catches, and where it fails — specifically in the context of Indian commercial contracts.

Clause Extraction: What AI Looks For

AI contract analysis tools scan for specific clause categories. Here are the most common ones relevant to Indian commercial agreements:

Open data/contract-clause-taxonomy.json in the code panel. This file contains 40+ clause types with their risk indicators, common positions (vendor-friendly vs buyer-friendly), and typical Indian market standards.

How Risk Scoring Works

Risk scoring assigns a numerical value to each clause based on how much it deviates from a "market standard" or "preferred" position. Here is a simplified example:

Example: Indemnity Clause Scoring

Clause: "Party A shall indemnify Party B against all losses,
damages, claims, and expenses arising from Party A's breach
of this Agreement, without any limitation."

Risk Analysis:
- Indemnity trigger: Breach of agreement ✓ (standard)
- Scope: All losses, damages, claims, expenses ✓ (broad but common)
- Cap: No limitation ⚠️ (HIGH RISK — uncapped indemnity)
- Carve-outs: None specified ⚠️ (MEDIUM RISK)
- Survival: Not specified ⚠️ (MEDIUM RISK — may survive indefinitely)

Risk Score: 8/10 (High Risk)
Recommendation: Add cap (typically 1x-2x contract value),
specify carve-outs, add survival period (12-24 months).

AI assigns these scores by comparing clause language against a database of thousands of reviewed contracts. The scoring is consistent — unlike a junior associate who might miss the uncapped indemnity on page 47 of a 60-page agreement at 11 PM before a deadline.

Before and After: Indian Commercial Contracts

Scenario 1: IT Services Agreement

Before AI review — A mid-size IT company in Bengaluru receives a 45-page Master Services Agreement from a US client. The legal team spends 3 days reviewing it manually. They catch the indemnity issue but miss that the governing law clause specifies Delaware law with exclusive jurisdiction in Delaware courts.

After AI review — AI scans the agreement in 4 minutes. It flags 12 risk items, ranked by severity. The governing law issue appears as Risk #2. The legal team spends 1 day on negotiation strategy instead of 3 days on review. Total time saved: 2 days per agreement.

Scenario 2: Vendor Onboarding

Before AI — A manufacturing company in Pune onboards 200 vendors annually. Each vendor agreement is reviewed by a single in-house counsel who takes 2-3 hours per contract. That is 400-600 hours per year on vendor agreements alone.

After AI — AI pre-screens all 200 contracts, flagging only the 35 that deviate from standard terms. Counsel reviews only the flagged contracts in detail. Time reduced from 500 hours to approximately 120 hours per year.

Open data/contract-risk-matrix.json to see a complete risk scoring matrix with 50+ clause types, their default risk weights, and Indian market benchmarks.

Indian Contract Law Essentials for AI Users

When using AI for contract analysis in India, you need to understand concepts that AI tools trained primarily on US/UK contracts may not handle well:

Stamp Duty and Registration

Under the Indian Stamp Act, 1899 (and state amendments), certain agreements must be stamped to be admissible as evidence. AI can flag whether an agreement type requires stamping, but the rates vary dramatically by state:

The Indian Contract Act, 1872

Key sections that AI should check against:

GST in Contracts

Every Indian commercial contract should address GST treatment. AI can check for:

Prompt Engineering for Contract Review

Here is how to use a general-purpose AI (like Claude) for contract review when you do not have a dedicated contract analysis tool:

Prompt: "Review the following contract clause and identify:
1. The type of clause (indemnity, liability, termination, etc.)
2. Which party it favours and why
3. Risk level (low/medium/high) with specific reasons
4. Indian law considerations (enforceability under Indian
   Contract Act, stamp duty implications, GST treatment)
5. Suggested amendments to balance the clause

Clause: [paste clause here]"

This structured prompt forces the AI to analyze systematically rather than give a vague summary. Always verify the AI's Indian law analysis against current statutes — AI models may not reflect the latest amendments.

Limitations of AI Contract Analysis

AI contract analysis has clear boundaries:

Key Takeaways