Digital Risk Control of PingAn P&C Qorus-NTT DATA Innovation in Insurance Awards 2026

1. Project Overview

The Chinese insurance industry faces an estimated annual fraud loss of at least RMB 50 billion. As a leading insurer with annual claims payouts exceeding RMB 200 billion, Ping An P&C proactively responds to national policy directives by leveraging cutting-edge technologies—including multi-modal AI engines and a Unified Risk Scoring Framework—to build a comprehensive "Omni-perception – Data-Driven Decision – Ecosystem Synergy" digital risk control system.

• Omni-perception: Efficiently harnesses unstructured data (e.g., images, voice recordings) to shift risk identification for tens of millions of annual cases from manual judgment to automated model discovery.

• Data-Driven Decision: Employs ensemble algorithms to construct a Unified Risk Scoring Framework, establishing a standardized "ruler" and common cognitive baseline for risk assessment. AI-driven comprehensive analysis ensures risk decisions are "faster and more accurate than human judgment."

• Ecosystem Synergy: Leveraging Ping An P&C’s proprietary data advantages, we collaborate with public security economic crime units and regulators to dismantle organized insurance fraud syndicates—leading industry-wide risk governance.

Furthermore, addressing the normalization of catastrophic risks, we pioneered a proactive, normalized management model—transforming disaster response from reactive to anticipatory. By integrating AI, IoT, and biometric technologies, we established an Intelligent Catastrophe Response System, integrating pre-disaster drills, real-time event perception, intelligent analysis, dynamic resource scheduling, and cross-departmental command. This system enables differentiated responses based on disaster scale and impact—achieving measurable risk reduction and minimizing customer property losses.

2. Project Background

The insurance industry confronts two systemic risk challenges:

Challenge A: Increasingly Sophisticated Insurance Fraud
Fraud tactics have evolved toward technological sophistication and professionalization. Fraudsters now deploy advanced AI tools—including voice synthesis to impersonate policyholders and image synthesis to forge medical records, accident scene photos, and official documents—rendering fraud increasingly covert. Organized, professional fraud syndicates—particularly in auto insurance—have formed complete black-market industrial chains spanning underwriting, staged accidents, vehicle repair, medical appraisal, and claims processing. These chains involve collusion among policyholders, repair shops, and medical institutions.

Traditional anti-fraud measures suffer from three critical pain points:

• Difficulty in Discovery: Clues are buried within massive volumes of heterogeneous unstructured data (e.g., synthetic voices, forged images). With tens of millions of cases annually, manual verification is operationally impossible.

• Slowness in Decision-Making: Quantifying risk probability and financial exposure requires multi-layered, multi-source analysis—typically involving three rounds of interdepartmental communication and taking 3–7 days.

• Weakness in Disposal: Traditional methods prioritize detection over evidence-based reasoning and forensic validation. In syndicated fraud contexts, insurers still rely primarily on isolated, case-by-case investigations.

Challenge B: Normalization of Major Disasters
Climate shifts—such as "southward movement of water resources" and "southward expansion of snow and rain"—have increased disaster frequency. Typhoons and rainstorms cause persistent high socioeconomic losses and escalating customer property damage. Sudden disasters can spike daily claims volumes to 3–6 times normal levels, easily overwhelming regional service capacities.

Ping An P&C handles one of the world’s largest claims portfolios: approximately RMB 200 billion annually (growing at ~20%), serving 500 million customer interactions across auto, property, accident/health, and agricultural insurance. Traditional claims operations are no longer sufficient to address these complex, dynamic risks. There is an urgent need to apply AI, IoT, and biometrics to precisely combat fraud, optimize cost control, and fulfill our core mission: safeguarding life and property while reshaping industry standards.

3. Competitive Landscape

This project establishes a sustainable competitive moat through three core technological innovations in risk perception—and as the first insurer to establish deep collaboration with China Banking and Insurance Information Technology Management Co., Ltd. (China BIIT), we co-built an industry-level AI anti-fraud engine, empowering the entire sector.

• Multi-modal Image Large Model (Internationally Leading): "Enabling Images to Speak" — unlocking latent insights from visual data. Its recognition accuracy significantly surpasses industry benchmarks (e.g., Tongyi Qwen, DS), and it was crowned Champion of ICDAR 2025 (International Conference on Document Analysis and Recognition).

• Vehicle Telematics Large Model (Industry First): "Making Vehicle Data Speak Human Language" — the first vertical large model and fault-code knowledge base built from scratch for telematics data. It translates obscure, non-standard vehicle signals into plain-language reports, revealing the true collision sequence.

• Non-Auto Risk Control Large Model (Industry First): "AI-ifying Expert Capabilities" — constructs granular risk profiles for 27,000+ non-auto products, combined with Chain-of-Thought (CoT) reasoning to automatically output risk types, evidentiary basis, contradictions, and conclusions—delivering a quantum leap in non-auto risk control capability.

4. Involved Departments

• Claims Operation Center

• Claims R&D Team, Technology Center

• Data Intelligence Platform Team, Technology Center

5. Key Achievements

• Fraud Prevention: Ping An P&C prevents over RMB 10 billion in fraud losses annually. Critically, AI-discovered and prevented losses now exceed those identified by human analysts. Losses detected solely by AI—undetected by humans, third parties, or rule-based systems, and resulting in claim denial—exceed RMB 1.1 billion and grow steadily year-on-year, providing a solid foundation for high-quality development.

• Industry Empowerment: We freely shared our Social Network Analysis (SNA) anti-fraud algorithms with the entire industry. Through our partnership with China BIIT, we co-built an industry-level AI anti-fraud engine covering 1 billion nodes and 10 billion relationships, identifying over 1,000 fraud syndicates annually, elevating sector-wide governance.

• Catastrophe Risk Reduction: Via the Intelligent Catastrophe Response System (pre-disaster warning/drills, rapid response, post-disaster review), we achieved an annual risk reduction of RMB 3.96 billion for customers despite increasing disaster frequency.