Insurers and Alibaba Cloud Discuss Trusted AI Deployment at 2026 Bund Summit

At the Insights Forum titled "Co-creating New Insurance Growth Services in the AI Era" at the 2026 Inclusion Bund Summit on September 11, 2026, industry experts and insurance executives from Alibaba Cloud, China Pacific Insurance and Ant Insurance discussed the trusted deployment of large models and AI agents in insurance, scenario innovation and the reshaping of user value, exploring a path toward high-quality digital and intelligent growth for the insurance industry.

Zhang Chi, vice president of Alibaba Cloud Intelligence Group, noted that the small, scattered AI agents previously used in the insurance industry generally suffered from problems such as being detached from business processes, unauditable and costly to iterate, making it difficult to meet the rigor required in finance. Alibaba Cloud has built financial trustworthiness engineering based on the Harness framework. This system accumulates industry scenario data, establishes standardized financial ontologies and compliant contract constraint mechanisms, addresses model drift through context and long-horizon task optimization, and sets up multi-layered compliance gates and closed-loop iteration mechanisms, turning AI from a probabilistic model into verifiable, auditable and accountable deterministic results.

Ye Mingman, assistant general manager of China Pacific Property Insurance, shared the insurer's AI practice. China Pacific Insurance has formulated a three-year AI+ development plan, integrating technology resources and computing platforms across the group to empower property insurance operations. On the service side, the company has built a service system centered on worry-free, reassuring and secure experiences, and proposed a "five direct and three optimized" intelligent service model for personal injury claims, under which AI agents can assist in five processes including direct dispatch of caregivers, direct delivery of rehabilitation services and direct assessment of injuries. On the risk-control side, China Pacific Insurance has built six intelligent risk-control systems covering fraud intelligence, liability determination, medical judgment, assessment evaluation, dispute mediation and quality testing. Leveraging collaboration between large and small models, these systems are driving risk control from post-event handling toward pre-event warning and in-process control.

Chen Guanhua, vice president of Ant Group and general manager of Ant Insurance, said the insurance industry faces information gaps between supply and demand, with products that are professional and complex and user needs that are vague and low-frequency, making it difficult for traditional manual and page-based service models to balance experience and efficiency. After three years of AI exploration, Ant Insurance has upgraded three industry AI services. "AI Insurance Selection Assistance" sorts out users' real needs through interaction, and click-through rates improve effectively after needs are calibrated. "AI Insurance Quick Search" covers more than 20,000 products, breaks down product differences in plain language and has served more than 14 million users cumulatively. "Yi Xiaobao," an AI insurance advisor upgraded to multimodal dialogue, can now provide proactive companion services and support report interpretation and voice interaction, with average conversations rising from four rounds to 8.6 rounds and interaction volume more than doubling.

Sun Zhenxing, chief technology officer of Ant Group's insurance business group, noted that insurance AI has entered a new stage of "business AI-ization," achieving self-learning and self-evolution in some business areas. Ant Insurance is exploring self-evolving systems and has achieved practical business results in document parsing, preliminary review of claims cases, assisted motor insurance purchase experiences and the AI insurance advisor "Yi Xiaobao." He said AI excels at tasks with clear experience and feedback that can be verified, while identifying key problems, system definition and emotional guidance still require human participation. Human-machine collaboration is one practical path for the industry, with the goal of translating AI technology into insurance services that users can perceive.