Zhixiaobang Qorus-NTT DATA Innovation in Insurance Awards 2026

Ping An, based on "AI in All" strategy, has developed ZXB (Zhixiaobang), its first intelligent sales support platform integrating multiple business scenarios. Through a unified intelligent agent entry point, the platform deeply integrates six core business scenarios—customer inquiry, product recommendation, customer visit, bidding & tendering, agricultural insurance, and enterprise health insurance—systematically addressing the three major challenges faced by frontline staff in their daily operations, enabling a transformation from "single-function tools" to an "intelligent collaborative system":

1. Multi-Agent Orchestration to Improve Issue Resolution Rates

   (1) Current Situation: Existing specialized intelligent agents only cover individual domains, with unclear capability boundaries. Users struggle to ask precise questions, resulting in low query resolution rates and fragmented user experiences.  

   (2) Solution: A unified intelligent agent entry is established, exposing only one smart application to users. The ZXB agent leverages intent recognition, business scenario understanding, and intelligent routing to deliver end-to-end service through a "one-query resolution" model. The system automatically identifies the domain of each query and dynamically dispatches the optimal specialized agent, enabling cross-domain collaboration and seamless integration.  

   (3) Data Validation: For example, frontline teams from Hunan’s group channel previously reported, “We cannot perform enterprise customer inquiries within the Intelligent Insurance Assistant.” After integrating with ZXB, the resolution rate for customer information-related queries increased from 20% to 84%.

2. Task Orchestration to Enable Multi-Agent Collaboration on Complex Issues

   (1) Current Situation: In complex business scenarios, the limited knowledge scope of a single specialized agent leads to fragmented responses. Users must manually piece together multiple answers, causing frequent operational interruptions, prolonged processing times, and poor user experience.  

   (2) Solution: An innovative "task management—task orchestration—task execution" triad technical framework is introduced. "Task management" categorizes similar Q&A into defined tasks, enhancing contextual understanding and memory retention. "Task orchestration" standardizes specialized agent capabilities into SOPs based on scenario identification, tool filtering, and permission control, enabling process-driven scheduling. "Task execution" supports follow-up on key factors, guidance for ambiguous queries, and multi-turn conversation storage and reasoning, improving closed-loop problem-solving capabilities.  

   (3) Data Validation: For instance, sales personnel in Hubei reported, “There is no intelligent way to recommend insurance products.” After adopting ZXB, the resolution rate for product recommendation queries rose from 60% to 80%.

3. Personalized Configuration for Customization and Rapid Deployment  

   (1) Current Situation: Different user groups (individuals, institutions, channels) have significantly varied needs. Developing custom intelligent agents from scratch for each group results in long development cycles, high costs, and low reusability. Additionally, localized underwriting policies and case examples are not included in current knowledge bases, making it difficult for frontline teams to accurately respond to regional customer demands.  

   (2) Solution: A visual configuration platform is built to allow on-demand assembly of specialized agent capabilities tailored to different user groups (individual, institutional, channel), enabling “plug-and-play, rapid deployment.” Furthermore, a dual-track knowledge base system combining “headquarters + local” data is innovatively adopted. Supported by a closed-loop mechanism of “data analysis—data annotation—bad case feedback,” the knowledge base is continuously iterated and updated, ensuring timeliness and accuracy.  

   (3) Data Validation: For example, local underwriters in Shanghai previously reported, “We cannot answer questions about coverage details of local products.” After integrating with ZXB, the resolution rate for location-specific knowledge queries increased from 72% to 86%.