Real engagements, real metrics
Knowledge base queryable in natural language
Knowledge Management Professional services · ~80 ppl
-68% search time
5→2 wks onboarding
Operational docs scattered across 4 systems; seniors burning 30 min/day looking up precedents. RAG indexing with source citations, Slack-integrated.
The client, an ~80-person Italian professional-services firm, had critical operational knowledge accumulated over the years and scattered across Google Drive, SharePoint, an on-prem legacy archive, and email. Senior partners burned on average 30 minutes a day looking for precedents, rulings and opinions previously issued on similar cases. New associates took 5-6 weeks to become operational.
We built a RAG knowledge base that indexes all documents from the 4 sources, normalises the metadata, and provides a natural-language search interface via Slack. Every answer includes source citations (link to document + paragraph) for verifiability.
Tech stack: pgvector for the vector store, OpenAI embeddings, Python application layer, Slack bot integration. Time to go-live: 4 months. Scale model: hand-off to the internal IT team + monthly care retainer.