1. Industry Background and Development Overview
1.1 AI Technology Application Status in the Insurance Industry
In the first half of 2025, the application of artificial intelligence technology in China’s insurance industry will undergo a qualitative leap, transitioning from conceptual exploration to comprehensive industrial implementation. According to the latest statistical data, by 2025, the coverage rate of AI technology applications in China’s life insurance industry has reached 92.6%, with a compound annual growth rate maintained at over 35%. The application of technology has moved from isolated breakthroughs to a stage of comprehensive integration. Against the backdrop of rapid development in financial technology and continuous improvement in regulatory policies, AI technology is accelerating the transformation and upgrading of the life insurance industry. AI technology has deeply integrated into the entire process of life insurance business, achieving significant results in key areas such as customer service, product innovation, and underwriting claims.
The deep development of AI applications in the insurance industry benefits from the continuous strengthening of the technological foundation. China Life Insurance has implemented the “Artificial Intelligence” action plan, with the frequency of AI capability invocation increasing by 27.2% compared to the beginning of the year, and the number of patent applications by the group has increased by 55.3% year-on-year.
Ping An has built a big data system covering finance, healthcare, and corporate operations, accumulating 30 trillion bytes of data, serving nearly 247 million individual customers. Zhong An Insurance has integrated mainstream domestic large models such as Tongyi Qianwen and Wenxin Yiyan, creating an AI middle platform architecture that combines multiple technologies.
1.2 Driving Factors for AI Applications in the Insurance Industry
The rapid development of AI applications in China’s insurance industry is mainly driven by three factors:
Policy Support:In March 2025, the Financial Regulatory Bureau, the Ministry of Science and Technology, and the National Development and Reform Commission jointly issued the “Implementation Plan for High-Quality Development of Science and Technology Finance in the Banking and Insurance Industries,” proposing to strengthen digital empowerment and encourage financial institutions to increase investment in digital transformation, utilizing technologies such as cloud computing, big data, artificial intelligence, machine learning, and privacy computing to develop digital business tools.For the insurance industry, the “Several Opinions of the State Council on Strengthening Supervision to Prevent Risks and Promote High-Quality Development of the Insurance Industry” clearly states that the use of artificial intelligence and big data technologies should be encouraged to enhance marketing services, risk management, and investment management levels.
Technological Progress: The maturity and application of generative AI technology, especially the explosive iteration of domestic open-source large models such as DeepSeek, have pushed the wave of “knowledge equality” in artificial intelligence to new heights. 2025 will be a turning point for AI applications in the insurance industry, where technical performance and costs reach a critical point, driving the industry from “extensive” scale-driven to “refined” value-driven transformation.
User Demand: Compared to industries such as internet retail and travel, the digital experience of users in the insurance field has long been lacking. Consumers are accustomed to instant, convenient, and highly personalized services in their daily lives, yet face complex terms, lengthy underwriting processes, difficult claims, and fragmented services in the insurance purchase and usage process. This gap in perception and experience has driven users to demand more intelligent, transparent, and interactive insurance services.
1.3 Value and Challenges of AI Applications in the Insurance Industry
According to McKinsey’s research, the insurance industry is expected to release productivity improvement potential of $50 billion to $70 billion driven by generative AI, covering the entire process from front-end sales and distribution to mid-end underwriting, policy management, and back-end claims, customer service, and middle and back-office operations.
At the same time, the application of AI technology in the insurance industry also faces multiple challenges:
Data Quality and Privacy Protection: Insurance institutions process vast amounts of data and complex transactions daily, making the accuracy and reliability of information crucial. If false information is fed into training models, it can contaminate the training data. Once deployed in risk analysis, underwriting claims, actuarial pricing, and asset allocation, it may trigger a chain reaction, causing immeasurable losses.
Algorithmic “Black Box” Issues: AI tools are trained based on past data, which may not accurately reflect reality or predict the future, leading to outputs that lack interpretability and are difficult to trace; biases in training data can result in unfair or erroneous algorithm outputs.
New Risk Prevention and Control: When most AI strategies adopt similar risk models, this homogenized response may accelerate the amplification of negative feedback loops, exacerbating market vulnerabilities.
Regulatory Compliance: Insurtech has deeply influenced the paradigm of insurance services, risk pricing mechanisms, and risk control models. Regulatory agencies need to adapt to new changes, closely monitor the application dynamics of AI technology in the industry, and enhance their ability to conduct in-depth analyses of intelligent algorithm risks.
2 AI Agent Applications and Technical Architecture in the Insurance Industry
2.1 Core Application Scenarios of AI Agent in the Insurance Industry
AI Agents have begun to demonstrate their empowering potential in key areas of the insurance industry, including customer service, risk control, and operational automation. From the perspective of the insurance business process, AI Agents have achieved large-scale applications in the following scenarios:
Pre-Sale Marketing and Customer Acquisition: AI Agents help insurance companies break through the bottlenecks of “difficult reach and slow conversion.” Corporate live broadcasts and digital humans serve as the front-end window to reach customers, breaking the limitations of time and space, making insurance products more vivid in explaining and conveying brand value. Marketing intelligent agents integrate marketing automation, corporate WeChat, SCRM, and event rights management to accurately outline 360° dynamic customer profiles, allowing insurance companies to clearly understand customers’ real needs.
Sales Team Empowerment : Intelligent training agents use the “Learn-Practice-Test-Assessment” full process to integrate champion sales scripts and real-world scenario simulations into daily training, helping new salespeople quickly master communication and closing skills. Marketing assistants combine top sales techniques with large model capabilities, providing real-time core product information and script guidance to sales personnel, making the entire process from customer acquisition to nurturing smoother.
Operational Transformation in Insurance: In the customer underwriting phase, OCR recognition quickly extracts underwriting and underwriting process information with financial-grade high precision, improving business efficiency. Multi-modal anti-counterfeiting based on large models and deep anti-counterfeiting technology accurately verifies customer identity and application materials, preventing fraud risks such as identity theft. Video dual recording, with leading RTC real-time audio and video capabilities, covers core links such as product promotion and risk warnings, combined with AI assistance to achieve deep anti-counterfeiting and intelligent quality inspection, improving first-pass approval rates and compliance.
Post-Sale Service Efficiency Improvement: When customers inquire or apply for claims, intelligent customer service covers customer contact centers, intelligent IVR, text robots, and other forms, providing 24/7 online service and quickly responding to diverse needs. When encountering complex issues, it automatically transfers to human customer service while synchronizing previous consultation records. Customers can also upgrade to 5G video customer service with one click, achieving real-time audio and video interaction in complex scenarios such as video inspections and injury visits, making services face-to-face.
Management Empowerment: At the enterprise management level, back-end support and front-end business need to form a linkage. Conversational BI integrates pre-sale, mid-sale, and post-sale business data, supporting natural language data Q&A and enhanced analysis, along with visual reports, providing managers with real-time and intuitive business analysis, greatly improving data analysis and application efficiency. Knowledge assistants effectively integrate internal and external knowledge resources, supporting multi-modal material analysis, summarizing massive fragmented knowledge into key Q&A points needed by employees, significantly improving office efficiency.
2.2 Technical Architecture and Core Capabilities of AI Agent in the Insurance Industry
The technical architecture of AI Agents in the insurance industry is undergoing significant transformation, with leading companies focusing on building an “AI middle platform” to achieve standardized output of technical capabilities. From a technical architecture perspective, it mainly includes the following layers:
Data Layer: Build an enterprise-level data architecture based on data lakes (such as Apache Iceberg), integrate data through ETL tools (such as DataX), and rely on data governance platforms (such as Apache Atlas) to achieve a data asset map. Based on data classification and grading, clarify data protection objects, and implement security management around data processing activities. For example, Ping An has databases covering finance, healthcare, and corporate operations, accumulating 30 trillion bytes of data, serving nearly 247 million individual customers.
Algorithm Layer: Access large models such as Tongyi Qianwen and DeepSeek, optimize insurance vertical knowledge responses through domain data fine-tuning and RAG (retrieval-augmented generation) technology, and achieve semantic matching with embedding models, integrating language, voice, image, and classification models, along with big data platforms and machine learning platforms, to provide stable support for AI applications.
Application Layer: Based on a unified AI middle platform, build AI Agent applications for different scenarios, including:
1. Multi-Agent Systems: The system adopts a hierarchical decision-making architecture, achieving automation of underwriting and claims through task decomposition and coordination algorithms (such as MARL). For example, Nuanwa Technology’s multi-agent system “Alamos” and “Lop Nur” provide AI underwriting and claims solutions, addressing the pain points of cumbersome processes and low conversion rates in traditional insurance, achieving full-process automation of underwriting.
2. Large Model Agents: For example, the “Shuidi Water Guardian” large model developed by Waterdrop Company has fully covered underwriting, customer service, and marketing processes. Its AI customer service “Baoxiao Hui” achieves 100% user coverage; the AI assistant “KEYI.AI” reduces the average processing time for complex health insurance underwriting by 80%, achieving an accuracy rate of 99.8% and a response speed increase of 260 times.
3. Intelligent Seat Assistants: China Life’s seat intelligent assistant supports a high connection rate for 95519, with the new version of the intelligent customer service robot achieving an accuracy rate of over 95%.
Interactive Layer: Integrate ASR (such as Alibaba Cloud Speech Recognition), TTS (such as Tencent Cloud Speech Synthesis), and multi-turn dialogue management engines (such as Rasa) to support real-time interaction with digital humans, enhancing user experience. For example, Zhong An Insurance’s “Lingxi Platform” has deployed nearly 110 intelligent robots, with a cumulative call volume of 450 million in the first half of the year, with each AI customer service managing over 60,000 users.
2.3 Evolution Path of AI Agent Technology in the Insurance Industry
The evolution of AI Agent technology in the insurance industry shows a trend from single functionality to multi-functional collaboration, from auxiliary decision-making to autonomous decision-making, and from closed systems to open ecosystems:
From Point Intelligence to Full-Scenario Digitalization: Early AI applications were mainly concentrated in single links, such as intelligent customer service or simple claims review. Now, leading insurance companies have achieved significant results through full-process intelligence coverage, with AI technology reshaping the insurance value chain in all links, promoting process optimization and efficiency improvement, driving transformation and upgrading in the insurance industry.
From Tool-Based to Decision-Making: AI Agents are transitioning from “empowering humans” to “providing human decision-making,” ultimately achieving a fundamental upgrade in operational models. Decision-making AI Agents adopt a hierarchical reinforcement learning (HRL) framework, combined with dynamic rule engines, gradually reducing human intervention. For example, Waterdrop Company launched the AI underwriting expert “KEYI.AI,” which, unlike traditional models relying on manual review, quickly processes health disclosures based on large model capabilities, intelligently matches underwriting rules, and provides more accurate risk assessments.
From Closed Systems to Open Ecosystems: With the technical support of API open platforms, insurance services are transitioning from closed to open. Open ecosystems standardize through RESTful API and GraphQL interfaces, integrating OAuth 2.0 authorization mechanisms to ensure secure data exchange. Data shows that this model has driven the penetration rate of UBI car insurance (insurance based on driving behavior) from 12% to 35%, and the conversion rate of embedded insurance on e-commerce platforms is three times that of traditional channels.
From “AI+Insurance” to “Insurance+AI” Transformation: The application of AI in the insurance industry has surpassed pilot and experimental stages, fully integrating into core links such as underwriting, claims, marketing, and customer service. This integration not only significantly enhances operational efficiency but also optimizes customer experience, making insurance services more personalized and imaginative while maintaining certainty.
3 Application Effectiveness and Case Analysis of AI Agents in the Insurance Industry
3.1 AI Agent Applications in Sales Assistance and Customer Service
In the fields of sales assistance and customer service, AI Agents have demonstrated significant application value:
Intelligent Customer Service Enhances Service Efficiency: China Life’s new intelligent customer service robot has an accuracy rate exceeding 95%, while Sunshine Insurance’s intelligent service satisfaction rate reaches 82%. China Pacific Insurance’s AI agents cover nearly half of the total customer service volume, with a consultation resolution rate stable at over 90%. Zhong An Insurance’s self-developed AI middle platform, “Lingxi Platform,” has over 100 active robots, with a cumulative call volume of 450 million in the first half of the year.
AI Underwriting Experts Improve Underwriting Efficiency: Waterdrop Company’s “KEYI.AI” underwriting system, based on a Transformer architecture large model, combines rule engines and knowledge graphs, trained on millions of historical underwriting data, achieving an accuracy rate of 99.8% and a response speed increase of 260 times.
Intelligent Outbound Calls Optimize Customer Acquisition Efficiency : Intelligent outbound agents rely on large model voice robots to achieve precise customer segmentation, while integrating social media leads for lead cleaning and nurturing, moving away from the inefficient “broad net” approach. A financial institution achieved a 50% increase in average call duration and a 30% increase in marketing conversion rate through intelligent outbound calls, optimizing customer acquisition efficiency and ROI.
Intelligent Training Enhances Sales Capability : The intelligent training platform has improved employees’ mastery of knowledge points by 50%, significantly accelerating team capability growth. For example, a leading financial institution adopted the intelligent training platform, resulting in a 50% increase in employees’ mastery of knowledge points and significantly accelerating team capability growth.
Case Study: Waterdrop Company AI Customer Service “Baoxiao Hui”
Waterdrop Company’s AI customer service “Baoxiao Hui” achieves 100% user coverage, covering multiple insurance customer service scenarios such as policy inquiries, policy management, and product Q&A, capable of accurately identifying user emotions, providing timely responses 24/7, and is expected to reduce the problem transfer rate by 50% and improve service efficiency by 50% throughout the year.
3.2 AI Agent Applications in Underwriting and Claims
Underwriting and claims are core links in the insurance business and are also the fields where AI Agent applications are most widespread and in-depth:
AI Underwriting Improves Review Efficiency: China Life’s digital underwriting system has improved the intelligent review rate to 95.8%, with the new version of intelligent customer service exceeding 95% accuracy. Sunshine Life’s AI underwriting system uses OCR technology to scan medical records, instantly matching millions of historical underwriting case data, providing conclusions of “standard underwriting” within 2 seconds, achieving a leap in efficiency from 3 days to 2 seconds, reconstructing the service logic of the insurance industry.
Intelligent Policy Issuance Improves Underwriting Efficiency: Ping An Property & Casualty Insurance uses multi-modal perception technology to successfully address the challenges of recognizing non-standard documents such as new car qualification certificates, achieving an intelligent policy issuance rate of 81.2%, with an average processing time reduced to under one minute. Zhong An Insurance’s cloud-based core insurance system “Wujieshan” issued 6.699 billion policies in the first half of 2025, with an automation rate of 99%.
Intelligent Claims Accelerate the Claims Process: China Ping An has created the “111 Fast Claims” new brand for life insurance claims, with a flash claim ratio of 59% in the first half of 2025. For complex medical documents such as medical records and admission/discharge records, it effectively breaks through the technical bottleneck of understanding accuracy, applying end-to-end automation in non-auto claims, covering nearly one million cases, achieving 55% automation in personal injury claims, with the fastest case closure time of 51 seconds.
AI Claims Decision Improves Accuracy: As of June 30, 2025, Nuanwa Technology’s AI claims solution achieved a maximum automatic review rate of 80%; in the first half of this year, it assisted insurance companies in reviewing approximately 2.3 million claims, with its fully automated claims review process achieving a decision accuracy rate of up to 98%.
3.3 AI Agent Applications in Risk Management and Cost Control
In risk management and cost control, AI Agents also demonstrate significant value:
Fraud Prevention and Risk Identification: Ping An Property & Casualty Insurance’s fraud prevention system intercepted losses of 6.44 billion yuan in the first half of the year, a year-on-year increase of 6%. China Pacific Insurance has established a vehicle insurance claims risk control tool based on image recognition, effectively identifying fraud risks and solving the problem of regional differences in claims, having detected risk amounts exceeding tens of millions.
Intelligent Risk Control Reduces Claim Rates: As of June 30, 2025, Nuanwa Technology has facilitated first-year premiums of 10.7 billion yuan, intercepting over one million high-risk insurance users, reducing claim rates by 10 to 23 percentage points. China Pacific Insurance’s vehicle insurance risk control tool based on image recognition has identified risk amounts exceeding tens of millions.
Cost Control Improves Operational Efficiency: China Pacific Insurance’s AI agents handle nearly half of the customer service workload, with health insurance claims automation reaching 16%, and the large model liability determination accuracy rate at 99%, reducing the cost per case by 47%.
4 Future Trends and Development Directions of AI Agent Applications in the Insurance Industry
4.1 Trends in Technological Integration and Innovative Applications
Multi-Modal AI Technology Becomes Industry Standard: With continuous breakthroughs in natural language processing, computer vision, and speech recognition technologies, insurance companies are striving to build an intelligent interaction system that covers all channels and scenarios through cross-modal attention mechanisms and feature alignment methods. Multi-modal AI achieves joint analysis of images (accident scene photos) and text (claims descriptions) through Vision-Language Transformers (such as VL-BERT), using cross-modal attention mechanisms to enhance loss assessment accuracy, significantly improving the level of intelligence in customer service. In practical applications, multi-modal AI not only optimizes traditional service processes but also creates a new model of intelligent service through human-machine collaboration, increasing service response speed by over 60% and significantly improving customer satisfaction.
Blockchain + AI Technology Integration Reshapes the Insurance Value Chain: Blockchain technology, with its decentralized and immutable characteristics, provides a trustworthy data storage environment for insurance business. When combined with the intelligent analytical capabilities of AI, it can build a more transparent and efficient business processing system. Especially in the claims service field, this technological integration achieves end-to-end automation from data rights confirmation to intelligent review, significantly enhancing business processing efficiency and credibility. Industry practices show that organizations adopting this technology have made breakthrough progress in claims timeliness and accuracy.
Deep Integration of IoT Technology and Health Management Opens New Spaces: Through health data collected from wearable devices and smart home terminals, significant support for insurance product innovation is provided. This data-driven innovation model not only achieves personalized customization of insurance solutions but also builds a virtuous cycle mechanism of “health promotion – risk prevention.” Market feedback indicates that these innovative products show strong momentum in customer acceptance and business growth.
Large Model-Driven Agent Evolution: Large models are accelerating the reconstruction of AI technology in insurance, enabling the industry to provide better risk management, insurance protection, and wealth management solutions, reducing operational costs, minimizing human errors, improving customer service, optimizing risk assessment, enhancing fraud detection, achieving automation in underwriting and claims, and innovating insurance product services. As large model technology continues to develop, AI Agents will possess stronger comprehension, reasoning, and creative abilities, bringing more innovative applications to the insurance industry.
4.2 Economic Benefit Analysis of AI Agent Applications
AI Agents in the insurance industry have brought significant economic benefits:
Cost Reduction: China Pacific Insurance’s AI agents handle nearly half of the customer service workload, with health insurance claims automation reaching 16%, and the large model liability determination accuracy rate at 99%, reducing the cost per case by 47%.
Efficiency Improvement: China Life’s digital underwriting system has improved the intelligent review rate to 95.8%, with the new version of intelligent customer service exceeding 95% accuracy. Sunshine Insurance has deepened the construction of intelligent customer service robots, gradually expanding the field of intelligent services, achieving a remote service automation rate of 65%, with intelligent service satisfaction reaching 82%.
Sales Growth: Waterdrop Company achieved a net operating income of 838 million yuan in the second quarter of 2025, with a net profit attributable to the parent company of 140 million yuan, achieving profitability for 14 consecutive quarters. Insurance-related income reached 739 million yuan, a year-on-year increase of 28.7%, with first-year premium scale reaching 3.204 billion yuan, a year-on-year increase of 80.2%.
Claim Rate Decrease: As of June 30, 2025, Nuanwa Technology has facilitated first-year premiums of 10.7 billion yuan, intercepting over one million high-risk insurance users, reducing claim rates by 10 to 23 percentage points.
5.3 Social Benefit Analysis of AI Agent Applications
AI Agents in the insurance industry have also brought significant social benefits:
Service Experience Improvement: China Life’s digital underwriting system has improved the intelligent review rate to 95.8%, with the new version of intelligent customer service exceeding 95% accuracy. Sunshine Insurance has deepened the construction of intelligent customer service robots, gradually expanding the field of intelligent services, achieving a remote service automation rate of 65%, with intelligent service satisfaction reaching 82%.
Inclusive Insurance Promotion: Under the policy orientation of “serving the real economy, preventing financial risks, promoting inclusive insurance, and enhancing people’s livelihood security,” the full-process digital product solution of Zhongguancun Science and Finance not only inherits the practical experience of leading insurance companies in AI full-link applications but also lowers the landing threshold for small and medium-sized insurance companies, accelerating the return of the insurance industry to “risk protection” fundamentals, making inclusive insurance more accessible and intelligent services closer to demand.
Customer Satisfaction Improvement: The customer service cost of virtual assistants in the insurance industry has been reduced by 35%, with NPS (Net Promoter Score) increasing by 20 percentage points, making “ask anytime, understand buying, and rest assured claims” a reality. Sunshine Insurance’s remote service automation processing rate has reached 65%, with intelligent service satisfaction reaching 82%.
Risk Reduction Management: China Ping An’s AI has empowered the enhancement of insurance risk control capabilities, strengthening abnormal behavior identification, intelligent risk assessment, and early warning. In the first half of 2025, Ping An Property & Casualty Insurance’s intelligent claims interception reduced losses by 6.44 billion yuan, a year-on-year increase of 6%.
5.4 Future Trends and Development Directions of AI Agent Applications
5.1 Technological Integration Trends and Innovative Applications
Multi-Modal AI Technology Becomes Industry Standard: With continuous breakthroughs in natural language processing, computer vision, and speech recognition technologies, insurance companies are striving to build an intelligent interaction system that covers all channels and scenarios through cross-modal attention mechanisms and feature alignment methods. Multi-modal AI achieves joint analysis of images (accident scene photos) and text (claims descriptions) through Vision-Language Transformers (such as VL-BERT), using cross-modal attention mechanisms to enhance loss assessment accuracy, significantly improving the level of intelligence in customer service. In practical applications, multi-modal AI not only optimizes traditional service processes but also creates a new model of intelligent service through human-machine collaboration, increasing service response speed by over 60% and significantly improving customer satisfaction.
Blockchain + AI Technology Integration Reshapes the Insurance Value Chain: Blockchain technology, with its decentralized and immutable characteristics, provides a trustworthy data storage environment for insurance business. When combined with the intelligent analytical capabilities of AI, it can build a more transparent and efficient business processing system. Especially in the claims service field, this technological integration achieves end-to-end automation from data rights confirmation to intelligent review, significantly enhancing business processing efficiency and credibility. Industry practices show that organizations adopting this technology have made breakthrough progress in claims timeliness and accuracy.
Deep Integration of IoT Technology and Health Management Opens New Spaces: Through health data collected from wearable devices and smart home terminals, significant support for insurance product innovation is provided. This data-driven innovation model not only achieves personalized customization of insurance solutions but also builds a virtuous cycle mechanism of “health promotion – risk prevention.” Market feedback indicates that these innovative products show strong momentum in customer acceptance and business growth.
Large Model-Driven Agent Evolution: Large models are accelerating the reconstruction of AI technology in insurance, enabling the industry to provide better risk management, insurance protection, and wealth management solutions, reducing operational costs, minimizing human errors, improving customer service, optimizing risk assessment, enhancing fraud detection, achieving automation in underwriting and claims, and innovating insurance product services. As large model technology continues to develop, AI Agents will possess stronger comprehension, reasoning, and creative abilities, bringing more innovative applications to the insurance industry.
5.2 Application Model Innovation and Business Transformation
From Insurance + to + Insurance Transformation: Insurance services will transition from standalone products to embedded services, integrating into people’s daily life scenarios. For example, when purchasing a mobile phone online, the system automatically pops up customized screen damage insurance based on your “screen damage history,” priced 40% lower than the generic version; when renting a car, AI recommends dynamic car insurance plans based on your driving score—this “when needed, it is there” scenario-based insurance is becoming mainstream. Behind this is the technical support of API open platforms: data shows that this model has driven the penetration rate of UBI car insurance (insurance based on driving behavior) from 12% to 35%, and the conversion rate of embedded insurance on e-commerce platforms is three times that of traditional channels.
From Passive Response to Proactive Prediction Service Model Transformation: AI Agents will shift from passively responding to customer needs to proactively predicting customer needs, providing forward-looking services. For example, when buying a health insurance policy, the accompanying service may not only be a compensation commitment but also a 24-hour online AI family doctor. It monitors your blood pressure and blood sugar data in real-time, automatically pushing health advice when abnormalities occur and assisting in scheduling specialist appointments, even directly connecting to hospitals for claims settlement when you are hospitalized—this is not a sci-fi scenario but a reality that Ping An Good Doctor has served 35 million users.
From Standardization to Personalized Product Design Transformation: AI technology will drive insurance products from standardization to personalization. Industry data shows that AI-driven personalized products are reshaping consumer experiences: in the health insurance field, the degree of customization based on dynamic data has increased by 60%, and the rate difference for cancer insurance can reach up to 5 times after incorporating genetic testing data; in the car insurance market, the degree of premium differentiation has increased by 40%, with more users discovering that “their premiums differ from their neighbors” is no longer a novelty. More importantly, this transformation allows insurance to shift from being a “luxury” to a “necessity”—when products truly match needs, consumer satisfaction significantly increases: the insurance purchase rate rises by 27%, and the renewal rate increases to 83%, with complaints about “buying the wrong insurance” and “insufficient coverage” decreasing by 52%.
From Single Service to Ecosystem Construction Transformation: Insurance institutions will transition from providing single insurance services to constructing insurance ecosystems. For example, Zhong An Insurance’s “Wujieshan” system not only supports the full-process management of insurance products but also shares data and capabilities with external partners, building an open insurance ecosystem. In the future, the insurance ecosystem will include health management, medical services, risk assessment, asset management, and other fields, providing customers with comprehensive risk management services.
5.3 Key Success Factors for AI Agent Applications in the Insurance Industry
Data Governance and Compliance Operations: Data governance and compliance operations are the foundational guarantees for AI applications. It is recommended that life insurance institutions focus on strengthening the protection of customer personal information, ensuring data security through encryption, access control, and other technical means. At the same time, a comprehensive lifecycle management mechanism for AI models should be established, conducting regular algorithm audits to ensure the compliance and fairness of technical applications.
Talent Team Building: Talent team building is a key factor supporting technological innovation. In light of the current shortage of composite talents in the industry, a diversified talent training system needs to be established. On one hand, insurance institutions can promote deep cooperation with universities and research institutions to jointly cultivate composite talents who understand both insurance business and AI technology; on the other hand, industry organizations can take the lead in establishing professional training bases and developing targeted curriculum systems. It is also important to improve talent incentive mechanisms, providing competitive development platforms for technical talents. According to industry estimates, by 2026, the demand for related talents will reach 150,000, requiring advance planning for talent reserve.
Organizational Management Innovation: Organizational management innovation is an important guarantee for promoting digital transformation. It is recommended that life insurance institutions establish dedicated digital transformation departments to coordinate the application strategies of AI Agents across various departments. In terms of organizational structure, more agile collaboration mechanisms can be attempted to break down traditional departmental barriers. Attention should be paid to the deep integration of traditional business personnel and technology teams to avoid the phenomenon of technology and business being “two separate entities.” Establishing a moderate innovation fault tolerance mechanism to provide necessary trial and error space for technical applications is also essential.
Industry Collaborative Development: Industry collaborative development is an effective way to enhance overall competitiveness. It is recommended to promote the establishment of industry technical application standard systems and formulate unified technical ethical norms. Building industry communication platforms to facilitate experience sharing and best practice promotion is also important. Under the premise of ensuring data security, exploring the establishment of industry-level data sharing mechanisms to promote the orderly flow of data elements is encouraged. Insurance institutions are encouraged to collaborate with technology companies and medical institutions to explore innovative application scenarios together.
5.4 Risks and Challenges and Response Strategies
Data Quality and Privacy Protection Risks: Insurance institutions process vast amounts of data and complex transactions daily, making the accuracy and reliability of information crucial. If false information is fed into training models, it can contaminate the training data. Once deployed in risk analysis, underwriting claims, actuarial pricing, and asset allocation, it may trigger a chain reaction, causing immeasurable losses. Response strategies include establishing a sound data quality management system, screening and cleaning external data sources to ensure data accuracy and credibility; promoting information interconnection in the financial and public service fields while ensuring security and privacy, releasing the potential of data elements; protecting user privacy through encryption and anonymization technologies; and establishing transparent data sharing mechanisms with users, allowing them to understand how their data will be used and giving them choices.
Algorithmic Bias and Fairness Risks: Biases in training data can lead to unfair or erroneous algorithm outputs. Strategies to address this include adopting fairness-aware machine learning algorithms (such as adversarial debiasing) and introducing SHAP interpretability tools for bias tracing. Response strategies include increasing algorithm transparency to ensure the interpretability of algorithm decision-making processes; establishing algorithm fairness assessment mechanisms to regularly check for biases; strengthening manual review and supervision to ensure the fairness and rationality of algorithm decisions; and establishing sound complaint handling mechanisms to promptly address disputes arising from algorithm decisions.
Model Security and Reliability Risks: AI-driven high-frequency trading enhances market response speed while amplifying volatility. When most AI strategies adopt similar risk models, this homogenized response may accelerate the amplification of negative feedback loops, exacerbating market vulnerabilities. Model security can be ensured through adversarial training and model watermarking technologies to prevent malicious attacks. Response strategies include strengthening model security testing and validation to ensure model stability and reliability; establishing model monitoring and early warning mechanisms to promptly detect and address model anomalies; implementing model version management to ensure controllability of model updates; and establishing model rollback mechanisms to quickly restore to previous stable versions when issues arise.
Regulatory Compliance Risks: Regulatory agencies need to adapt to new changes, closely monitor the application dynamics of AI technology in the industry, and enhance their ability to conduct in-depth analyses of intelligent algorithm risks, improving the rules and systems for intelligent algorithms, enhancing algorithm interpretability, transparency, fairness, and security, ensuring precise, moderate, flexible, and efficient regulation. Response strategies include strengthening communication with regulatory agencies to stay informed about regulatory policies and requirements; establishing sound compliance management systems to ensure AI applications meet regulatory requirements; conducting regular compliance self-inspections and audits to promptly identify and rectify compliance issues; and participating in industry self-regulatory organizations to jointly promote the normative development of the industry.
6. Conclusion and Recommendations
6.1 Core Value and Development Prospects of AI Agent Applications in the Insurance Industry
AI Agents in the insurance industry have transitioned from point intelligence to full-scenario digitalization, from auxiliary decision-making to autonomous decision-making, and from closed systems to open ecosystems. AI technology is reshaping the insurance value chain in all links, promoting process optimization and efficiency improvement, driving transformation and upgrading in the insurance industry.
Core Value: The core value brought by AI Agents to the insurance industry includes: enhancing operational efficiency, reducing costs; improving risk identification capabilities, lowering claim rates; improving customer experience, enhancing customer stickiness; promoting product innovation, meeting personalized needs; driving business model transformation, and expanding market space.
Development Prospects: With the continuous advancement of AI technology and the deepening digital transformation of the insurance industry, the application prospects of AI Agents are broad. In the future, AI Agents will achieve deep applications in more scenarios, such as actuarial pricing, investment decision-making, product design, and other core links, bringing more innovation and transformation to the insurance industry. It is expected that by 2026, the original premium income of China’s insurance market will exceed 6.3 trillion yuan, with life insurance, property insurance, and health insurance all showing growth trends; the insurance industry’s technology investment will continue to grow rapidly, with total investment expected to exceed 67 billion yuan in 2025, with big data, cloud, and AI as the main investment areas.
6.2 Strategic Recommendations for AI Agent Applications in Insurance Institutions
Strategic Positioning and Planning: Clearly define the strategic positioning of AI Agents: Elevate the application of AI Agents to a strategic level, clarify the positioning and goals of AI Agents in the overall strategy of the company, and develop a clear application roadmap for AI Agents. For example, China Pacific Insurance has clearly proposed three core strategies: health management, internationalization, and “artificial intelligence,” building an enterprise-level AI capability system.
5. Strengthen Top-Level Design: Establish a leadership team for AI Agent application involving senior leaders to coordinate and promote the application strategies of AI Agents across various departments, ensuring that AI Agent applications align with the overall strategy of the company.
6. Set Phase Goals: Based on the actual situation of the company, set short-term, medium-term, and long-term goals for AI Agent applications, ensuring steady progress and effectiveness in AI Agent applications.
Technical Architecture and Capability Building:
7. Build AI Middle Platform: Establish a unified AI middle platform, integrating internal and external data and AI capabilities, providing standardized and modular AI services for various business departments, lowering the application threshold for AI and improving application efficiency. For example, Zhong An Insurance has accessed mainstream domestic large models such as Tongyi Qianwen, DeepSeek, and Wenxin Yiyan, building a solid data platform, integrating core technologies such as embedding models and distilled large models, and combining various models such as language, voice, image, and classification, along with big data platforms and machine learning platforms, to provide stable support for AI applications.
8. Strengthen Data Governance: Establish a sound data governance system to improve data quality and availability, providing a solid data foundation for AI Agent applications. For example, China Ping An has built a big data system covering finance, healthcare, and corporate operations, accumulating 30 trillion bytes of data, serving nearly 247 million individual customers.
9. Train AI Talents: Strengthen the training and introduction of AI talents, establishing a team of composite talents with knowledge of both AI technology and insurance business to provide talent support for AI Agent applications. For example, China Life has implemented the “Artificial Intelligence +” action plan, with the frequency of AI capability invocation increasing by 27.2% compared to the beginning of the year, and the number of patent applications by the group increasing by 55.3% year-on-year.
Application Implementation and Operational Management:
10. Scenario-Driven: Focus on high-value and high-potential application scenarios driven by business scenarios, prioritizing the implementation of AI Agent application projects that can quickly generate value, such as intelligent customer service, intelligent underwriting, and intelligent claims. For example, China Ping An has created the “111 Fast Claims” new brand for life insurance claims, with a flash claim ratio of 59% in the first half of 2025; for complex medical documents such as medical records and admission/discharge records, it effectively breaks through the technical bottleneck of understanding accuracy, applying end-to-end automation in non-auto claims, covering nearly one million cases, achieving 55% automation in personal injury claims, with the fastest case closure time of 51 seconds.
11. Agile Development: Adopt agile development methods for rapid iteration and continuous optimization, ensuring that AI Agent applications can respond promptly to changes in business needs and improve application effectiveness.
12. Human-Machine Collaboration: Establish a human-machine collaborative working model, fully leveraging the advantages of AI Agents and human experts to achieve complementary strengths, improving decision quality and efficiency. For example, in the underwriting phase, AI Agents can handle standardized and clearly defined underwriting tasks, while complex and high-risk underwriting tasks are handled by human experts.
13. Continuous Optimization: Establish evaluation and optimization mechanisms for AI Agent applications, regularly assessing application effectiveness, promptly identifying issues, and optimizing to ensure that AI Agent applications continue to generate value.
6.3 Industry Ecosystem Construction and Win-Win Cooperation
Industry Collaboration:
14. Promote the Formulation of Industry Standards: Participate in the formulation of industry standards for AI Agent applications, promoting the normative development of the industry.
15. Strengthen Industry Communication: Actively participate in industry communication activities, share experiences, learn advanced practices, and jointly promote industry development. For example, build industry communication platforms to facilitate experience sharing and best practice promotion.
16. Establish Data Sharing Mechanisms: Explore the establishment of industry-level data sharing mechanisms to promote the orderly flow of data elements while ensuring data security.
Cross-Industry Cooperation:
17. Cooperation with Technology Companies: Collaborate with technology companies that lead in AI technology to jointly develop AI Agent applications suitable for the insurance industry, accelerating technological innovation and application landing.
18. Cooperation with Medical Institutions: Collaborate with medical institutions to obtain medical data, improving the accuracy and efficiency of risk assessment, and developing more targeted insurance products and services.
19. Cooperation with Regulatory Agencies: Actively communicate with regulatory agencies to understand regulatory policies and requirements, jointly promoting the compliant application of AI Agents in the insurance industry.
As AI technology continues to advance and the digital transformation of the insurance industry deepens, the application prospects of AI Agents are broad. Insurance institutions should seize opportunities, actively embrace change, and use AI Agent applications as a lever to promote digital transformation and innovative development, gaining advantages in fierce market competition.