
Inspiration Legal compliance is a massive bottleneck for businesses. Companies spend millions manually reviewing contracts for prohibited terms like "unlimited liability" or "perpetual duration." We wanted to prove that AI Agents can do more than just read these contracts—they can actively fix them. What it does ComplianceGuard is a multi-agent system that acts as an autonomous legal defense team. * RiskScanner (The Auditor): Powered by llama-3-2-90b-vision-instruct, this agent analyzes legal text and consults a Knowledge Base of corporate policies to identify violations (RAG). ComplianceFixer : When a risk is detected, the Scanner autonomously delegates the task to this specialized agent, which rewrites the clause to be safe and compliant. How we built it We built the entire solution on IBM watsonx Orchestrate. * Orchestration: We linked two distinct agents to enable collaborative delegation. * Reasoning: We utilized ReAct logic so the agents "think" and plan before acting. * Knowledge Integration: We uploaded specific corporate policy documents that the agents retrieve to validate specific liability caps and termination clauses. Accomplishments that we're proud of We successfully implemented an Agentic Workflow where one AI commands another. The system is fully autonomous moving from detecting the risk to rewriting the final legal clause without human intervention.
23 Nov 2025