AuditEase

Inspiration

The global economy loses over $4 trillion annually to non-compliance. But for us, the inspiration was more specific: realizing that highly paid lawyers spend thousands of hours manually hunting for "needles in haystacks." We saw companies like Sephora getting hit with massive fines ($1.2M) simply because a vendor contract was missing a single privacy clause. We realized that contract review is currently a "ticking time bomb"—it’s slow, human-dependent, and opaque. We wanted to build a system that turns legal compliance from a bottleneck into a competitive advantage. We asked ourselves: What if an AI could not just read a contract, but act as a Senior Risk Auditor that quantifies risk in actual dollars?

What it does

AuditEase AI is an agentic compliance platform that transforms static legal documents into verified knowledge assets. Instead of vague "risk scores," it calculates Financial Liability (e.g., "$5.0M Exposure"), translating legal jargon into business impact. Users upload a contract (PDF/DOCX) and select a regulatory standard (GDPR, HIPAA, SOC2). The AI analyzes the document in under 60 seconds. It detects non-compliant clauses and automatically drafts a Legally Compliant Rewrite that users can apply with one click. A RAG-based "AI Consultant" allows users to ask specific questions (e.g., "Does this contract allow data transfer to the US?") and get answers based strictly on the document's context.

How we built it

We built AuditEase as a modern, high-performance Single Page Application (SPA).

AI Engine: We leveraged Google Gemini 1.5 Flash via the Generative AI SDK. Its massive context window allows us to ingest entire agreements without truncation, and its reasoning capabilities power our liability calculations.

Frontend: Built with React (Vite) and TypeScript for type safety. We used Tailwind CSS and Shadcn/UI to create the "Interstellar" glassmorphism aesthetic.

Backend & Database: We used Supabase for secure authentication and PostgreSQL storage. We implemented Row Level Security (RLS) to ensure strict data isolation between users.

Visualization: We integrated Recharts and Plotly to render dynamic risk gauges and liability trend lines.

Challenges we ran into

Getting a Large Language Model (LLM) to output consistent, parseable JSON for our dashboard (risk scores, dollar amounts, clause IDs) was difficult. We spent significant time refining our system prompts to enforce strict schema adherence so the frontend wouldn't break. Early versions of the AI would sometimes invent regulations. We solved this by implementing a strict "Grounding" prompt strategy, forcing the model to cite specific articles from the uploaded Regulation PDF before making a claim. The animated "moving network" background and glassmorphism effects initially caused lag. We optimized the CSS animations and used lightweight SVGs to maintain a smooth 60fps experience.

Accomplishments that we're proud of

We successfully engineered a prompt chain that can logic out a realistic financial penalty based on regulatory fine structures (e.g., GDPR's 4% global turnover rule). We built a seamless "Diff View" where users can see the original dangerous clause side-by-side with the AI's safe rewrite is a huge UX win. We didn't just build a hackathon prototype; we built a product that looks and feels like expensive SaaS software, complete with a document vault, search filtering, and tiered pricing logic.

What we learned

We learned that treating English prompts like code—with strict syntax, logic gates, and error handling—is essential for building reliable AI Agents. The quality of the audit depends entirely on how well we structure the input context (the regulation text vs. the contract text). Lawyers are used to terrible software. Giving them a sleek, "Interstellar" interface makes the complex task of compliance feel manageable and even engaging.

What's next for AuditEase

Multi-Agent Negotiation: We plan to build a feature where two AI agents (one representing the buyer, one the seller) can autonomously negotiate standard clauses to reach a middle ground.

Slack/Email Integration: Allowing users to forward a contract to [email protected] and receive a risk report back in their inbox.

Built With

• postgresql
• react
• sonner
• supabase
• toaster
• typescript
• vitest