The challenge
Call-center agents at a telecommunications provider had no efficient access to internal knowledge spread across Confluence documentation and past Jira tickets. Incoming email-based incident tickets required manual classification, routing and resolution lookup, causing slow response times and high agent workload.
At a glance
StackRAG platform, enterprise LLM, vector embeddings, Confluence & Jira APIs
DeploymentOn-premise Linux
Duration~3 months to go-live
What we did
SWICON deployed an on-premise RAG system in the client's Linux environment, with automated ingestion of the Confluence knowledge base and Jira tickets via API, initial bulk migration and continuous incremental sync. Semantic vector embeddings drive retrieval; an enterprise LLM generates answers. AI-based text classification routes incoming email tickets and proposes solutions for fault reports. Agents work from a web UI, with a reporting dashboard for management.
The result
A prototype was in agents' hands at week 6 and the system went live at week 13, with automated ticket classification and routing, resolution suggestions for fault reports and a continuously synchronised knowledge base.
