FinanceAIRAGKnowledge BaseChina

Pacific Insurance Private Knowledge Base Intelligent Q&A System

Project Background

As a leading insurance company in China, Pacific Insurance possesses a massive volume of business documents, including insurance clauses, claims rules, and product materials, totaling over 30,000 documents. Claims adjusters, sales consultants, and new hires need to frequently consult these documents daily to answer customer inquiries and perform business operations. Traditional keyword searches struggle to accurately locate information; employees spent an average of 15 minutes finding required content, and new employee training cycles lasted 3–6 months, severely impacting operational efficiency and customer experience.

Core Pain Points

Low knowledge retrieval efficiency: Over 30,000 documents relied on keyword search, with an average retrieval time of 15 minutes and accuracy below 40%.

Long training cycles for new hires: New employees required 3–6 months to handle tasks independently, leading to high training costs.

Asynchronous knowledge updates: Frequent changes in clauses and rules made it difficult for employees to obtain the latest information in a timely manner.

Data security and compliance requirements: Insurance business data involves customer privacy, prohibiting the use of public cloud AI services.

Solution

Private RAG Architecture Deployment

Deploy a RAG (Retrieval-Augmented Generation) system based on Qwen2.5-72B within the intranet, performing structured chunking and vector indexing on over 30,000 documents to build an enterprise-level knowledge graph. All data processing and inference are conducted within the intranet, meeting the financial industry's compliance requirements for data not leaving the domain.

Intelligent Q&A and Knowledge Recommendations

Implemented a natural language Q&A interface, supporting multi-turn conversations and contextual understanding. The system not only returns precise answers but also provides original clause citations and related knowledge recommendations, helping users fully understand business rules. Average retrieval time reduced from 15 minutes to 10 seconds.

Automated Knowledge Base Updates

Integrated with the company's content management system, automatically triggering incremental index updates when clauses or rules change, ensuring the knowledge base is always synchronized with the latest business rules and eliminating the risk of information delays.

Effect Data

Metric	Before	After	Improvement
Knowledge Retrieval Time	15 min	10 sec	99%
Answer Accuracy	40%	92%	130%
New Employee Training Cycle	3–6 months	1–2 months	67%
Knowledge Update Delay	7 days	Real-time	100%

Tech Stack

Qwen2.5-72B, Milvus Vector Database, LangChain, FastAPI, Vue.js, Nginx Intranet Deployment

“After the knowledge base went live, the claims team's work efficiency doubled. Most importantly, all data stays on the intranet, fully meeting compliance requirements.”