Back to Portfolio
Revenue OperationsAIAutomation

Multi-Step Contract Intelligence & RAG Platform

We built an AI-powered contract analysis platform that lets business users ask complex, multi-part questions about their customer contracts using plain English. By combining OpenAI reasoning, dynamic tool routing, and a fully managed retrieval pipeline, the platform analyzes contracts and tabular data at scale — no SQL or data science required. Users upload new documents or CSVs, instantly create embeddings and database tables, and run queries that chain up to 15 backend actions — all through a simple React interface. The result: contract intelligence and analytics delivered in minutes, not weeks, without hiring a team of analysts.

An AI agent tool to chat with documents and databases

Industry

Legal

Duration

3 weeks

Technologies Used

React logo
Reactfull-stack
React is a front-end JavaScript library used for building user interfaces, especially single-page applications. It’s used to create interactive UI components and manage how a web app looks and behaves in the browser.
OpenAI logo
OpenAIai-ml
OpenAI is an artificial intelligence (AI) research and deployment company.
PostgreSQL logo
PostgreSQLdatabase
PostgreSQL, often called Postgres, is a powerful and popular open-source relational database management system
A person with a brain made of computer chip
Retrieval Augmented Generationai-ml
Retrieval-Augmented Generation makes language models reference a knowledge base outside of its training data sources before generating a response.

The Challenge

Contract review is slow, expensive, and typically locked behind legacy tools or technical gatekeepers. The client needed a way for business users — not just data teams — to run deep-dive, compound queries across contracts, PDFs, and tabular data, with instant ingestion and zero SQL required.

Our Approach

We architected a multi-tool reasoning agent that interprets each user query, chooses between standard LLM response, SQL tool call, or hybrid vector search — or chains all three for complex questions. Users can upload new documents or tables, trigger instant chunking/embedding, and map data to new or existing tables. The React UI is wizard-based and non-technical, designed for speed and self-service.

Our Solution

  • Natural Language Querying: Users describe what they want, the agent orchestrates everything — SQL, vector search, and multi-step workflows.
  • Rapid Data Ingestion: Upload PDFs, DOCX, CSV, or pasted text; wizard walks users through table/column mapping and embedding creation.
  • Hybrid Search: Combines SQL for structured queries and vector search for semantic clause/document retrieval.Scalable Orchestration: Up to 15 sequential tool calls to handle layered, compound queries and analytics.
  • Explainable Results: Every answer comes with transparent reasoning and links to the original contract source.

Results

  • •Enabled business users to answer legal/operational questions in minutes
  • •Unified all contract, document, and tabular data for cross-source querying
  • •Reduced contract analysis cycle time by 80%