Internal knowledge base that actually gets used

The Challenge

Policy documents, compliance guides, and product specifications were scattered across SharePoint, shared drives, and individual inboxes. New hires took months to become productive, and even experienced staff spent significant time searching for the right document or asking colleagues.

The Solution

A regional insurance company had policy documents, compliance guides, and product specs scattered across SharePoint, shared drives, and people's inboxes. New hires took months to get up to speed. We built a domain-specific AI assistant fine-tuned on their internal corpus — about 12,000 documents. The key wasn't just the AI; it was building a good retrieval layer so the assistant cites its sources. Adoption hit 78% within the first month because people could trust the answers.

Technical Approach

We built a RAG-based AI assistant fine-tuned on the company's 12,000-document internal corpus using DeepSeek. The retrieval layer uses ChromaDB with semantic chunking and source citation, integrated via SharePoint API for real-time document sync. The key to adoption was transparent source attribution — every answer links back to the original document.

Technology Stack