AI-Powered Contract Review for Hong Kong Law Firm
How a mid-size law firm cut contract review time by 70% and saved HK$200K+ annually — with zero data leaving their premises.
Client Overview
Client Type
Mid-size Hong Kong law firm (12 lawyers)
Hardware
Mac Studio M4 Max · 48GB unified memory · 1TB SSD
Challenge
Manual contract review taking 4–6 hours, privacy concerns with cloud AI
Result
70% faster review, HK$200K+ annual savings, fully private
The Challenge
Hours lost to manual review
The firm's lawyers were spending 4 to 6 hours per contract on manual review — identifying key clauses, checking for non-standard terms, and cross-referencing against previous agreements.
With client confidentiality as a non-negotiable priority, sending contracts to cloud-based AI services was out of the question. The firm needed a solution that kept all data on-premises while dramatically reducing review time.
The Solution
Local AI contract intelligence
We configured a Mac Studio M4 Max with PrivateGPT and a custom RAG pipeline. The firm ingested over 5,000 contracts into a local Qdrant vector database, creating a searchable knowledge base of their entire contract history.
- Clause extraction and comparison in seconds
- Non-standard term detection with risk scoring
- Cross-reference against 5,000+ historical contracts
- Bilingual support for English and Chinese contracts
Results
Measurable impact
Our clients trust us with their most sensitive deals. Local AI means that trust is never broken.
— Managing Partner, Mid-size HK Law FirmTech Stack
Tools powering this solution
Ollama
Local LLM runtime for contract analysis and clause extraction
PrivateGPT
Document ingestion and RAG pipeline for contract Q&A
Qdrant
Vector database storing 5,000+ contract embeddings locally
Custom Embeddings
Fine-tuned embedding model optimised for legal terminology
Hardware Used