Chao Yan
Product Leader, 360AI Enterprise Knowledge Base
Head of the AI Knowledge Base Product in the Digital Intelligence Product Department at 360. With 9 years of experience in enterprise services, collaborative office solutions, and enterprise-level intelligent applications, they have extensive expertise in product planning and implementation across areas such as unstructured data management, enterprise knowledge management, document security, document intelligence, and large-model-based knowledge Q\&A. As an early pioneer in applying large language models in the industry, they have helped the team achieve commercial breakthroughs in sectors such as energy, manufacturing, higher education, and government, with over 10 successfully deployed "Knowledge Management + AI" projects.
Topic
From Document Piling to Scenario Breakthrough: Building Enterprise Knowledge Bases in the Era of Large Language Models
Topic Overview: In the era of large AI models, teams often fall into the inefficient trap of "document piling," which fundamentally stems from a lack of systematic knowledge governance. Based on practical experience, this talk reveals the logic behind knowledge base reconstruction: 1. Breaking Cognitive Misconceptions: Using the DIKW pyramid to analyze the transformation from data → information → knowledge → wisdom, emphasizing that the core of knowledge governance is “scenario-driven decision support.” 2. Capability Framework: * Knowledge Construction: Deriving knowledge needs from business scenarios, avoiding information overload through multi-source integration and structured organization; * Knowledge Retrieval: Achieving precise recall via vector search, text segmentation, knowledge metadata, and knowledge graphs, addressing RAG challenges; * Knowledge Updating: Establishing a closed-loop evolution mechanism with user feedback, system analysis, and scenario evaluation; * Knowledge Security: Building a security system around knowledge storage, processing, sharing, and AI agent applications. 3. Platform Implementation Practice: Breaking down the six-layer architecture of knowledge governance platforms (data collection → metadata → knowledge construction → knowledge base → retrieval → application), illustrating how knowledge lifecycle management is embedded into product workflows. 4. Product Manager Role Upgrade: Transforming from a “requirements collector” to a “knowledge architect,” leading scenario breakdown, knowledge modeling, and iteration, enabling large models and knowledge bases to become efficiency-boosting “digital employees.”
