Turning scattered knowledge into an assistant for support
Our client already had the answers to its customers’ questions, but they were scattered across old tickets, documentation, FAQs and blog articles. We centralised that knowledge and built an assistant that uses it to help the support team.
- Client
- Customer support
- Industry
- Customer support
- Outcome
- 4 sources brought together in one assistant
Context
For confidentiality reasons, this client remains anonymous.
Over time, a support team builds up valuable knowledge: answers in the ticket history, documentation, FAQs, published articles.
That knowledge exists, but it is hard to use: each source has its own format and tool, and finding the right information takes time.
The challenge
The company had extensive knowledge spread across ticket histories, documentation, FAQs and blog articles.
The challenge: make that knowledge easy for support teams to use.
Our approach
A general-purpose AI assistant knows neither your products nor your procedures. To be useful, it has to answer from the company’s actual knowledge, not from general knowledge.
What we built
The data foundation
We centralised and structured the four knowledge sources into a foundation the AI can use.
Historical tickets, documentation, FAQs and existing content are now organised around a single search system that finds information by meaning, not just keywords (semantic search).
The AI layer
We developed a RAG-based assistant: before generating a response, it identifies the relevant information in the knowledge base.
The assistant therefore answers from what the company actually knows.
Results
Knowledge previously scattered across four sources has become a single foundation the AI can use.
Support teams rely on a context-aware assistant, fed by the company’s history and documentation.
TechnologiesRAG · LLM · Knowledge Base · Semantic Search · AI Assistant