Ping An Health Insurance launches AI claims and cancer-care agents

Ping An Health Insurance has released two AI agents, one for claims processing and one for managing cancer care, as Chinese insurers push large language models from pilots into core service operations. The company framed the launch as part of parent Ping An Group's “Service Year” initiative and a step toward using AI at scale to improve efficiency and customer experience.

The claims agent, called “AI Claims Expert,” is designed to handle the full claims workflow. A single claim can require organizing and checking medical records, invoices and itemized expenses, then judging liability against policy terms through information verification, liability assessment, cost calculation and conclusion review. Ping An says the agent covers more than 1,000 claims checkpoints and combines claims specialists' experience with medical knowledge, business rules and policy language, improving through expert feedback and case reviews.

It supports both fully automated and human-AI collaborative modes. Cases with complete materials, clear information and unambiguous liability can be closed automatically after quality checks, which the company says happens with 99.7% accuracy. Cases needing more documents or confirmation move to a claims workbench, where the agent presents open items as interactive cards and staff make the final judgment with AI analysis alongside.

In a pilot covering pediatric medical products, Ping An says average time from submission to closure fell from days to hours, with the fastest case closed in three seconds, and claims operations efficiency rose about 80%. The insurer had already applied AI agents to claim reporting, document review and payouts, reaching automation rates above 70% for information collection and above 60% for review without human handling.

Ping An plans to extend the claims agent to more than 1,000 individual insurance products and more than 10,000 group products. Claims are both a cost center and the moment customers judge whether insurance actually works, so the test for such systems is whether faster automation can preserve accuracy and traceability at the same time.

The second agent, “AI Specialty Disease Manager,” was developed with Ping An Technology for cancer patients. It offers report interpretation, medical consultation, treatment-path advice and rehabilitation management across the course of illness, covering breast, lung, gastric and cervical cancers. Ping An says it handles about 120,000 tumor-related claims customers a year, and that roughly 80% of treatment plans are adjusted after multidisciplinary consultations.

Before diagnosis the agent provides report interpretation, condition assessment, suggested tests and staging; during treatment it offers multidisciplinary plan options, comparisons and referral guidance; afterward it handles side-effect management, rehabilitation follow-up, nutrition and psychological support. It also provides round-the-clock medical consultation and patient records. Its knowledge base includes more than 100,000 oncology cases, an evidence library at the tens-of-millions scale, more than 2,000 cancer hospitals and a database of over 60,000 oncologists.

The company says the underlying model ranked first on the HealthBench Hard evaluation and passed certification from the China Academy of Information and Communications Technology, with 91% accuracy for treatment plans and a 95% adoption rate for rehabilitation plans. The agent will be integrated into Ping An's health services and case-management workbench. Extending insurance from post-hoc payouts into continuous care is an industry-wide ambition, but how AI recommendations fit with physicians' responsibility and how outcomes are verified over time still need more public evidence.