A legal assistant that searches the entire archive
Lawyers spent hours searching for clauses and precedents across thousands of contracts and documents. The knowledge lived in people's heads and in folders, hard to find and impossible to scale.
We built a RAG assistant over the internal archive: documents are indexed in a vector database, and the Claude model answers questions citing the exact source and page. Access is role-controlled, and the data never leaves the client's infrastructure.
As the archive grows, the assistant becomes ever more valuable, with no additional organising effort. Lawyers spend more time on substantive work and less on searching, and new colleagues become productive much faster, with instant access to the firm's accumulated knowledge.
- Anthropic Claude
- RAG (embeddings)
- pgvector / PostgreSQL
- Python
- OCR
- RBAC
Measured impact
“I find in 10 seconds what used to take me half a day.”
