From the initial camera survey to the subsequent application of the claims system, and now to the mobile survey, the insurance company spent 10 years and only realized the electronicization of accident photos and claims data.

But this year, big changes happened.

Four companies, Ant Financial, Ping An of China, PICC and PICC, have entered the field of loss assessment from different angles with the help of technology.

Among them, Ant Financial and Ping An of China focus on the efficiency of loss assessment, and the newly launched intelligent loss assessment tool aims to greatly improve the efficiency of simple claim settlement.

When the efficiency of damage assessment in simple cases reaches the second level, there will really be no need to wait for simple car repairs in the future.

The People's Insurance Company of China has developed and purchased a variety of accident vehicle detection tools. If you want to know the damage of the vehicle without disassembly and inspection, the focus is on the quality of claims settlement.

China Baoxin relies on the industry's most complete database to focus on anti-fraud, and the probability of insurance fraud is expected to be reduced.

Now, this intelligent loss determination competition has begun. Who will have the last laugh?

The battle between loss-defining treasure and AI intelligent flash compensation. Ant Financial is the first giant to announce this year that it will empower artificial intelligence to determine losses.

The company’s launch of Dingshengbao in June gave it another identity - a technology provider for the insurance industry.

Jiemian News learned that Dingshengbao uses artificial intelligence to simulate the manual work process in the auto insurance loss assessment process to help insurance companies realize automatic loss assessment for simple cases.

Behind the scenes is the application of image recognition technology in the insurance field. AI is used as the eyes and brain of the loss assessor, using algorithms to identify accident photos and determine damaged parts and the extent of the loss.

Then, by connecting with the repair price library of the cooperative insurance company, damaged parts, repair plans and repair prices can be provided within a few seconds.

This is the first time that image loss assessment technology has been commercially used in the field of auto insurance.

However, this loss-assessment tool, which claims to "revolutionize the lives of 100,000 claims settlement people," has been questioned by many in the claims field since its launch.

The first point of doubt lies in the massive amount of historical image data.

To achieve the effect of image recognition and damage assessment, machine learning is required from massive historical car insurance accident photos.

"70% of the premium income of the auto insurance industry is in the hands of the three major property insurance giants PICC, Ping An and China Pacific Insurance, which means that about 70% of insurance claims data is in the hands of these large companies," said Wei Wei, founder of A6 Studio, a senior claims adjuster.

However, these are the core data of insurance companies. All three companies are capable of data analysis, and it is unlikely that they will be shared with Ant Financial.

In fact, PICC, Ping An and China Pacific Insurance are all promoting artificial intelligence.

An insider from Ant Financial’s insurance business group told Jiemian News that a large number of historical auto insurance photos are indeed needed to train the machine. Dingshangbao’s machine learning partners are large and medium-sized insurance companies such as Dadi Property Insurance and Taiping Property Insurance, which have enough historical data to feed the machine.

study.

This person further added, "Different insurance companies have different car models and photo shooting specifications. Some companies take more flat shots, some take more overhead shots, some emphasize detail shooting, and some like to make annotations. We use various

Photos from various insurance companies are 'fed' to machine learning, so that Dingshengbao can cover all car models and adapt to photos taken in different weather and different shooting methods."

Another issue that has been questioned more is the localization of accessories prices.

The most important part of determining the damage is to determine the repair price, which involves the price of accessories, auxiliary materials and labor hours in various regions across the country.

As an Internet financial company, Ant Financial obviously does not have this data.

In this regard, the above-mentioned insider of Ant Financial’s insurance business group said, “In essence, Dingshengbao does not involve this part. Dingshengbao only does the picture recognition and loss determination in the previous stage, and the system will automatically submit the results to the insurance company. Pricing

The power is still in the hands of the insurance companies, and each insurance company determines the price based on its own price database and outputs the maintenance plan."

In addition to technical doubts, the industry is most worried about moral hazard issues.

Some senior claims adjusters interviewed by Jiemian News generally said that they have not seen the anti-fraud function of Dingshengbao and are worried that insurance fraudsters will use this loophole to bring risks to insurance companies' risk control.

According to Jiemian News reporter, after discussions with cooperative insurance companies, Dingshengbao has some anti-fraud designs on the logical relationship between accessories and the authenticity of accidents, which are used to judge the authenticity of accidents and provide early warning for some items that do not conform to the logic of accessories loss.

case.

However, the above-mentioned people from Ant Financial’s insurance business group also admitted that currently only some simple anti-fraud measures can be done.

“We will continue to deepen in this direction in the future. This is a process of continuous optimization,” he said.

Ant Financial’s monopoly on the field of intelligent loss determination did not last long amid skepticism.

Last month, Ping An of China announced the launch of its own smart loss assessment product - AI Smart Flash Compensation, which is the first smart loss assessment tool launched by an insurance institution.

Similar to Dingshengbao, AI smart flash compensation also uses image recognition technology and machine learning to train on massive historical real claim image data to achieve image recognition and loss determination.

Jiemian News learned that "it has a large amount of historical car damage picture data" is the main advantage of this part.

Ping An of China has a 19.83% property and casualty insurance market share, which means it has about 20% of the industry’s vehicle damage data.

Ping An of China publicly stated that AI intelligent flash compensation is currently the only artificial intelligence loss assessment and risk control product that has been put into use in real production environments.