How Siemens Harnesses Generative AI to Build the Enterprise Knowledge Chatbot “XiaoYu”

Siemens China’s IT DA team faced fragmented internal information, limited knowledge‑sharing capabilities, and a lack of knowledge‑operation tools, prompting the development of an integrated enterprise knowledge solution called “XiaoYu”.

The project began with a clear business need: to improve employee productivity and innovation efficiency by leveraging large language models (LLMs) for internal knowledge retrieval and interaction.

“XiaoYu” is built on a Retrieval‑Augmented Generation (RAG) architecture. All corporate documents are vectorized and stored in a vector database; when a query arrives, it is also vectorized, matched against the stored vectors, and the relevant context is fed to the LLM to generate the final answer.

Key technical challenges included optimal document chunking (balancing response speed and coverage), multimodal data integration (using OCR to extract text from images), and ensuring answer accuracy through intent‑recognition models that route queries either to the vector store or to a generic knowledge base.

The system also provides reference material alongside generated answers, allowing users to verify information. To improve usability, the chatbot was integrated into web portals, mobile apps (iOS and Android), and offered voice output via TTS.

Since its launch in May, the platform has attracted over 17,000 users—about half of Siemens’ workforce—with roughly one‑third of queries related to Siemens business topics. It supports sales teams by delivering customer background, contract history, and real‑time bidding information, and it automates sales‑visit record creation via voice input.

An IT service robot built on the same technology offers 24/7 support for common IT issues and includes an automated monitoring system (“Eagle Eye”) that detects batch‑processing failures and triggers alerts without human intervention, covering over 1,000 devices and 200 applications.

Beyond the chatbot, Siemens has constructed a broader “DaYu” ecosystem that layers data integration, AI services (LLM, CV, recommendation, NLP), and self‑service platforms, enabling eight product lines across finance, logistics, approval workflows, and sales to benefit from standardized data and AI capabilities.

The article emphasizes strategic considerations such as ROI, avoiding POC traps, and the importance of aligning AI initiatives with business value, talent development, and organizational change.

Overall, the case illustrates how a large enterprise can rapidly adopt generative AI, combine it with robust data infrastructure, and deliver tangible productivity gains across multiple business domains.