Mobility company with 50+ technicians and hundreds of procedure manuals. They needed technicians to find the right answer in seconds.
Sector: Mobility · Technical services · movicat.cat ↗
Technicians had a shared folder with 1,200 procedure PDFs. When an incident arose, they spent 10-15 minutes looking for the right procedure (and often used an outdated one).
We built a semantic AI knowledge base:
PHP + Python (backend), MySQL + ChromaDB (vectors), Anthropic Claude (LLM), OpenAI ada-002 embeddings, responsive web app.
Without calling a senior colleague.
Natural question → answer with cited source.
New technicians autonomous in 2 weeks instead of 2 months.
Strict RAG: only answers with source; otherwise "I don't know".
System detects recurring unanswered questions and flags what to document.
Every answer logged: who asked, what was answered, which source.
We cut incident resolution time by 35%. New technicians are autonomous in weeks, not months.
— Operations director
Perfect case for semantic AI. Typical cost: €3,500-7,000 + ~€30/mo model. ROI in 2-4 months.