ClauseIQ – AI-Powered Multi-Agent Contract Risk Analyzer

Transform contract review from hours to minutes with intelligent multi-agent AI

---

🧾 Problem Statement

Contract review is slow, tedious, and expensive. Organizations signing vendor agreements, SaaS licenses, NDAs, and partnership contracts must identify legal and financial risks before approval — but this requires time, legal expertise, and manual effort.

Reading each clause word-by-word, identifying obligations, renewal traps, liability exposure, confidentiality terms, and negotiating leverage is mentally exhausting.

The Result:

• Teams spend hours reviewing documents, or
• Sign contracts without fully understanding their exposure
• Leading to harmful financial commitments, unlimited liabilities, and operational risks

Manual review makes it impossible to scale: more contracts means more delays, more risk, and increased workload for legal and procurement teams.

---

💡 Solution

ClauseIQ applies multi-agent AI to automate contract understanding. A user uploads a PDF, and the system performs three core steps automatically:

1. Extract clauses from the contract and structure them into categories
2. Evaluate legal and financial risks
3. Generate an executive summary targeted at business decision-makers

Users can also ask follow-up questions such as:

• "Which clauses are high risk?"
• "What should we renegotiate?"
• "Explain the indemnification clause"

ClauseIQ stores memory across the session, enabling grounded follow-up answers based strictly on the uploaded contract rather than hallucination. This turns contract review into a fast, scalable workflow that empowers decision-makers with clarity before signing.

---

🏗 Architecture

ClauseIQ is a multi-agent system built using Google's Agent Development Kit (ADK). It is not a single general model — instead, it is a team of specialized agents coordinated by a central orchestrator.

Agent	Responsibility
Extractor Agent	Reads PDF artifacts and extracts structured clauses
Risk Analyst Agent	Evaluates risk level and business/legal exposure for each clause
Reporter Agent	Writes an executive-style summary for business leadership
Root Agent	Manages delegation and answers follow-up questions

🔹 Extractor Agent

Reads the uploaded PDF using ADK's artifact loader and returns structured JSON with:

• Clause ID
• Title
• Content
• Category (payment, termination, liability, IP, confidentiality, renewal, indemnification, warranty, general)

🔹 Risk Analyst Agent

Analyzes each clause to determine risk level (low, medium, high) with justification — such as:

• Unlimited liability
• Auto-renewal
• Penalty clauses
• Perpetual IP rights

Provides recommendations for each risk identified.

🔹 Reporter Agent

Synthesizes clause and risk data into a concise executive summary written for non-legal business audiences. It highlights risk distribution and negotiation talking points.

🔹 Root Agent

Acts as the user-facing coordinator. It:

• Launches the pipeline automatically after PDF upload
• Enables follow-up conversations
• Leverages ADK memory, tools, and session state to avoid repeated processing
• Grounds answers in the uploaded contract

---

🛠 Tools & Utilities

ClauseIQ uses three ADK tools to support specialized reasoning:

Tool	Purpose
load_artifacts_tool	Reads the uploaded contract PDF
load_memory_tool	Remembers extracted clauses, risks, and summaries across queries
google_search_tool	Optional external factual research if needed

By combining tools with specialized agents, ClauseIQ converts:

Contract text → Structured clauses → Risk intelligence → Business-ready insights

---

📋 Technical Implementation

Model Configuration

• Model: Gemini 2.5 Flash Lite
• Retry Configuration: 5 attempts with exponential backoff
• HTTP Status Codes Handled: 429, 500, 503, 504

Agent Definitions

1. Extractor Agent

• Output: JSON array of clauses
• Categories: payment, termination, liability, IP, confidentiality, renewal, indemnification, warranty, general
• Tools: artifacts_tool, memory_tool

2. Risk Analyst Agent

• Output: JSON array with risk assessments
• Risk Levels: low, medium, high
• Analysis Factors: One-sided terms, unlimited liability/indemnity, automatic renewal, penalties, perpetual IP rights
• Tools: memory_tool, search_tool

3. Reporter Agent

• Output: Executive summary
• Report Structure:
  • Overall risk rating
  • Contract overview (4-6 sentences)
  • Top critical issues (3-5 bullets)
  • Negotiation recommendations/redlines
• Tools: memory_tool

4. Root Agent

• Role: User-facing coordinator
• Workflow:
  1. Call extractor_agent to obtain clauses
  2. Call risk_analyst_agent to assess risks
  3. Call reporter_agent to generate summary
• Tools: artifacts_tool, memory_tool, AgentTools for all sub-agents

---

🔧 Setup Instructions

Requirements

pip install google-adk google-genai pymupdf

Environment Configuration

import os

os.environ["GOOGLE_API_KEY"] = "your_api_key"
os.environ["GOOGLE_GENAI_USE_VERTEXAI"] = "FALSE"

Creating the Agent

adk create clauseiq_agent --model gemini-2.5-flash-lite --api_key $GOOGLE_API_KEY

Running ADK Web Interface

adk web --url_prefix {url_prefix}

---

📊 Usage Example

1. Upload a contract PDF through ADK Web interface
2. Ask: "Analyze this contract and summarize the risks"
3. Follow-up queries:
   • "Which clauses are high risk and why?"
   • "What should we renegotiate before signing?"
   • "Explain the indemnification clause"

The system maintains session memory, so follow-up questions are answered based on the already-processed contract data without re-extraction.

---

🚀 Value Statement

ClauseIQ reduces contract review time from hours to minutes while improving clarity and risk awareness. It empowers founders, procurement teams, and legal-ops professionals to make confident decisions without requiring a full legal review for every document.

Key Benefits

• ⏱ Time Savings – Review contracts in minutes instead of hours
• 📉 Risk Reduction – Identify hidden liabilities automatically
• 📈 Scalability – Handle more contracts without increasing headcount
• 🧠 Intelligence – Data-driven insights for better negotiation

---

🔮 Future Enhancements

With more time, the next evolution would add a Negotiation Agent to automatically recommend redlines and generate counter-proposal text — turning contract analysis and contract negotiation into a unified AI-driven workflow.

---

📌 Conclusion

ClauseIQ demonstrates how multi-agent systems can transform legal and business processes. Instead of relying on a single general model, ClauseIQ divides the complex task of contract review into focused subtasks handled by dedicated agents — extraction, risk evaluation, and executive summarization — coordinated by a Root Agent.

Running on ADK Web, the system supports interactive review: users upload a contract, receive immediate insights, and ask follow-up questions without reprocessing the PDF. This modular architecture makes the system scalable, reusable, and easy to extend.

---

📄 License

This project is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

---

👨‍💻 Author

Nitin Dave

• Kaggle: @nitindave

---

🔗 Quick Links

Resource	Link
📓 Kaggle Notebook	ClauseIQ Final Version
🎥 YouTube Demo	Watch Demo
📝 Kaggle Writeup	Competition Entry
🏆 Competition Track	Enterprise Agents – Agents Intensive Capstone Project