Contents
Technical writers often need to search large collections of source documents to find precise, complete and reliable information for manuals, service content and other technical documentation. At kothes, we developed an internal AI chatbot based on Retrieval-Augmented Generation (RAG) to support this research process. The system receives user questions, searches a document pool for relevant information and prepares answers in a chat interface.
However, we quickly learned that standard RAG approaches are not sufficient for the high requirements of technical communication: answers must be accurate, traceable and complete. In this talk, we share our journey from the first prototype to a more robust solution. We explain the editorial challenges, the technological hurdles, the research-based methods we investigated, and what the chatbot can do today.
Takeaways
Attendees learn why basic RAG is often not enough for technical writing and how retrieval quality, answer precision and completeness can be improved in practice.
Prior knowledge
Basic knowledge of AI and RAG is helpful, but not required.