Inspiration
Businesses review thousands of contracts every year, including vendor agreements, master service agreements, partnership agreements, licensing agreements, and procurement contracts. These documents are often lengthy, complex, and require significant manual effort to review.
While exploring real-world contracts filed with the SEC, I realized that critical clauses such as indemnification, termination rights, renewal provisions, confidentiality obligations, and intellectual property ownership are often buried within dozens or hundreds of pages of legal text. Missing these clauses can expose organizations to financial, operational, and legal risks.
This inspired me to build ClauseWatch, an AI-powered contract analysis agent that helps users quickly understand contracts, identify risks, and extract actionable insights.
What it does
ClauseWatch allows users to upload a contract and automatically:
Extract key contract information Identify important clauses Detect potential legal and business risks Generate executive summaries Highlight obligations and responsibilities Surface renewal and termination provisions Analyze indemnification and liability clauses Provide confidence scores and supporting evidence
Instead of spending hours manually reviewing documents, users can obtain meaningful insights within seconds.
How we built it
The system combines Large Language Models with structured contract analysis workflows.
Workflow User uploads a contract Contract text is extracted and processed AI analyzes the document section by section Key clauses are identified and classified Risks are detected and scored Supporting evidence is linked back to contract sections A structured summary is generated for the user Technologies Used Google Gemini React TypeScript Vite Node.js Modern AI prompting techniques Real-world SEC contract datasets
The application was tested on real contracts including:
Master Service Agreements Joint Venture Agreements Promotion Agreements Transaction Advisory Agreements Commercial Partnership Contracts
Challenges we ran into
One of the biggest challenges was reducing hallucinations and ensuring that every detected risk was supported by actual contract language.
Many legal clauses are written differently across contracts even when they have the same meaning. For example, indemnification, liability limitations, and termination rights can appear in significantly different formats depending on the industry and contract type.
Another challenge was prioritizing meaningful risks. Not every unusual clause is necessarily a critical business risk. We focused on surfacing issues that legal, procurement, and business teams would genuinely care about.
Finally, working with real-world contracts revealed how complex legal documents can be. Many agreements exceed dozens of pages and contain cross-references, amendments, schedules, and attachments that must be interpreted together.
What we learned
Through this project, I gained experience in:
Building AI agents for document intelligence Prompt engineering for legal text analysis Contract risk assessment workflows Retrieval and evidence-based AI outputs Frontend and backend integration Working with large real-world legal datasets
Most importantly, I learned that AI is most valuable when it augments human decision-making rather than replacing it. ClauseWatch is designed to help users review contracts faster while keeping humans in control of final decisions.
What's next for ClauseWatch
Multi-contract comparison Clause benchmarking against industry standards Contract obligation tracking Renewal and compliance reminders Team collaboration features Custom risk policies for enterprises Contract repository search and analytics
Log in or sign up for Devpost to join the conversation.