China In-Depth: The impact of data analytics and AI on China’s insurance industry
Although AI has become the inevitable answer to a range of distribution, product development, and operational questions, the ability to extract insights from new data streams does have exciting and far-reaching consequences for the industry. However, the depth of data held by insurance companies is marred by the fact that insurers have yet to apply AI based technologies to it. Considering this, and the emergence of several new AI use cases, we review the most interesting of these initiatives; from new product development, claims management, and distribution.
- New product development
The prospect of accessing new data formats such as location data from smartphones, health metrics from wearables, and online user behavior in order to develop highly tailored and usage based insurance products has long been awaited. However, using such data to upgrade decades-old actuarial models has proven more difficult in practice. First, data quality from hardware varies significantly between devices and sensors. Second, public opinion towards data privacy has turned significantly towards conservatism. Finally, China’s recently merged insurance and banking regulator has renewed scrutiny of insurer’s pricing models and the factors underpinning them. Notwithstanding the challenges, several Chinese insurers and startups have re-imagined the use of AI across health and personal accident lines.
First, Ping An has become China’s largest private health insurer by entering a 75/25 joint venture with South Africa’s Discovery Group. By doing so, Ping An gained access to Discovery’s pioneering product-development methodologies such as establishing a ‘single view’ customer database that enables Ping An to tailor and then bundle financial services to its 80m active customers.
Elsewhere, CPIC has been testing usage-based auto insurance policies. However, although recently liberalised, factoring in location and driving behavior data is still prohibited and several Chinese insurtechs have chosen to pilot new product lines in neighboring Asian markets. One of these is Wukongbao, a Beijing based broker that has used advanced analytics and AI driven product recommendations to develop an API that monitors user behaviour and then prompts personalized personal accident insurance to users. By working with 5,000 third party platforms including travel aggregators, healthcare portals, auto marketplaces, ride-sharing apps, and niche hobbyist apps, Wukongbao’s API has enabled it to be chosen ahead of hundreds of traditional insurers and brokers who could otherwise sell similar products through these third parties.
Ultimately, although the advent of smartphones and advances in sensor technologies has unlocked new data formats, the availability and accuracy of such data remains a challenge, and regulatory approval of new pricing models is desperately needed before a new generation of usage-based products can emerge from China.
- Anti-fraud Initiatives
China suffers one of the highest insurance fraud rates globally. It is estimated that 25% of claims are fraudulent and the practice is particularly widespread in auto, health and credit insurance with China’s state owned auto insurers are particularly plagued by fraudulent claims with little visibility into repair shops.
Facing this, and seeking to evolve state owned incumbents, the government has turned to a tried and tested solution; the private sector. By partnering with several analytics startups 4Paradigm, which we have already featured here, is working with China’s biggest property insurer, PICC, to identify fraudulent claims by 4S repair centers in 600 cities across China.
Second, China’s near-universal basic health insurance scheme is heavily subsidised by the government, which leaves it open to exploitation by public hospitals that provide excessive treatments and even fabricate medical services in the knowledge that insurers have little visibility into these practices.
Recognising this, Hangzhou based Tongdun has developed anti-theft and fraud management software for a host of internet insurers such as Zhong An and An Xin. Tongdun, which styles itself as China’s answer to Palantir, has established a unified database, connecting public hospitals, medical institutions and insurance companies so that they can share patients’ records, standardise medical services and allow insurers to precisely estimate business costs.
Finally, compared with policy-based auditing, machine learning algorithms can simultaneously identify a variety of fraud indicators to reduce unreasonable claims. Here, customer authentication, CRM and security services are all being reimagined. Again, Ping An has set the pace with a ‘biometric authentication’ feature that offers policy holders the option to enhance their account access through face and voice recognition.
The feature, delivered through Ping An’s ecosystem of apps, scans the structure of the face (especially the nose and eyes) with a higher degree of accuracy than the human eye and is also being used to verify the identity of insurance applicants and on-board new agents.
- Loss Adjusting
A separate but equally important element of claims processing is loss adjusting. In this case, China’s internet giants are contributing with AI-based loss adjusting services. Of these, Alibaba’s is the most sophisticated and its white label image recognition service has been adopted by several property insurers to assess the exterior damage of auto insurance claims. Although making hard earned innovations publicly available to competing insurers may come as a surprise given the intensity of competition in China, the fact that this system requires billions of images to assess damage, produce repair estimates and validate claims means that making the platform publicly available will extend the lead that Ant has established.
Additionally, this is a radically different method of processing auto claims, it’s worth noting that even the most sophisticated loss adjusting solution has limitations – for example, internal damage that is invisible to external photography or prior-generation models that look similar but have structural differences. Ultimately, Alibaba’s contribution is a good example of an AI-based solution that reaches deeper into the value chain than purely distribution focused efforts.
- Ping An
Of all the efforts mentioned above, one company is steering a course through each of them; Ping An.
Founded in 1988, Ping An first pioneered the use of analytics by unifying data collected by its offline agents with its outbound telemarketing and its online platform. At the heart of Ping An’s efforts to become an analytics-driven insurer is its Pinnacle Lab based in Shenzhen where it stores data its AI is trained on.
Ping An also has the upper hand in terms of access to data which it uses to form a relationship with pending policy holders (those who have applied for life/health insurance but are waiting for the policy to be activated). Since launch, the biometric identification feature has lowed the rate of customers who withdraw their application during this window from the industry average of 4% to 1.4%.
Ping An HealthKonnect is working on extracting original information from a patient’s medical history, then parse, clean, transform and export them into standardised data from their cloud-based platform. In partnering with more than 200 cities throughout the country, Ping An is helping provide them with services ranging from cost control and actuarial services to medical resource management and health record applications. Although initially intended purely to maximise the renewals and cross selling rate, Ping An’s advanced analytics quickly morphed into a company wide data-driven operation. Although Ping An has set the standard for the use of analytics and AI, many others are embracing AI driven user experiences to optimise distribution, automate claims and expedite operations.
The volume, variety and velocity of data available to insurance companies is increasing at a dizzying pace. The examples above demonstrate the wide ranging efforts currently underway in China, and, with government support and unmatched scale, China has the opportunity to lead the development of AI-based technologies globally. However, the fact that insurers have yet to capitalise on their data is an irony not lost on the industry. Additionally, amid a relentless quest for new technologies, a bigger challenge is acquiring talent, as AI development takes a high degree of judgement to turn insights derived from analytics into action.