tigerhiddendragonswordofdestiny| What are the opportunities for small and medium-sized insurance companies in the insurance industry to "overtake" with the intensive implementation of AI models?

Interface News reporter | Lu Wenqi

"I am 30 years old with an annual income of 200000 yuan. If you want to buy deposit insurance, please recommend it for me." Interface News tried to update the Chat-Trust3 released by Xinmei Mutual Life Insurance on the 22nd.Tigerhiddendragonswordofdestiny.0 ask a question like this. Subsequently, Chat-Trust3.0 quickly generated a detailed demonstration of insurance interests and a summary of the characteristics of a certain annuity product.

As a data-intensive industry, insurance is one of the best application fields of artificial intelligence model. Since the AI model craze was set off by ChatGPT at the end of 2022, many insurance companies, such as PICC, Ping an of China, Sunshine Insurance, Zhongan Insurance and so on, have laid out the layout around the big model.

In the interview, a number of technical people analyzed the interface news that the popularity of the AI model will reduce the gap between large insurance enterprises and small and medium-sized insurance enterprises, and with the lower and lower price of the general large model, small and medium-sized insurance enterprises have the opportunity to greatly improve their operational efficiency. However, AI technology still has hallucinations, data security and other problems, and there is still a long way to go to reshape the industry.

AI narrowing the gap between insurance companies

The combination of insurance and large model is inseparable from the rapid development of large model in this year. Compared with the tens of millions of dollars of pre-training, in the era of universal large model ability to rush forward, using general large model to build vertical application has become a more feasible method for insurance companies to enter the big model.

In 2023, PICC released the "Mathematical Smart-PICC Big Model" and launched two PICC exclusive Q & A model applications-"PICC Zhiyou" and "Congmingbao", which are for individual and corporate customers respectively. Provide insurance product consultation, claims application, insurance policy management and other services, aimed at improving user experience and enterprise insurance management efficiency.

Sunshine Insurance also listed the self-developed AI model as the company's strategic project in 2023, and its Sunshine Zhengyan GPT model has been applied to customer service, sales support and other scenarios.

Tong Guohong, head of Xinmei Mutual Life data Information Center, told Interface News that the input of the big model mainly comes from two parts, one is the input of arithmetic, the other is the input of the technical team. "in terms of computing power, as a small and medium-sized insurance company, our cost input is more focused on the company's existing rigid needs, and then expand step by step, but we believe that with the further development of technology, the cost of computing will gradually decline."

Wang Boyang, an artificial intelligence investor in Jiacheng Capital, told Interface News that with the development of AI technology, the gap between small and medium-sized insurance companies and large insurance companies will be narrowed. "maybe we think that large insurance companies will have more data and resources and use big models earlier. But now the general large model is becoming more and more intelligent, and the cost is gradually reduced. A series of core business of the insurance industry, such as intelligent customer service and pricing data, can be assisted by the big model, so as to help small and medium-sized insurance enterprises improve their business efficiency and achieve business breakthroughs. "

Zeng Yu, head of Xinmei's digital operations department, said that through the AI model, the cost of insurance companies could be reduced by 99 per cent ideally. If insurance companies can master this technology, the savings can be invested in better customer service and lower-priced products, which will bring great advantages and create insurance companies with a new organizational structure.

Caitong Securities Research News believes that, on the one hand, the terms of insurance products are complex and there are many professional terms, which puts forward far higher requirements for investment in the recruitment and continuous training of sales staff of insurance companies; on the other hand, back-end services such as underwriting and indemnification also need a large number of personnel docking, so the insurance industry is generally regarded as one of the manpower-intensive industries with high labor costs. And under the background of stricter supervision, the policy side has more stringent professional requirements for insurance employees, and insurance companies have the active demand for scientific and technological empowerment.

AI large model is mainly used in auxiliary work.

So, what are the possible combinations of AI and insurance?Tigerhiddendragonswordofdestiny?

Yang Fan, chairman of Xinmei Mutual, said that the development opportunities and application scenarios of the big model in the insurance industry are very broad, such as providing personalized content and product recommendations for different customer groups; by analyzing historical and real-time data, help insurance companies more accurately assess risks and set insurance rates; speed up the two-core process and enhance customer service experience.

However, he also pointed out that at this stage, the insurance industry AI large model is limited by the ability of the basic large model, and it takes a large amount of money to achieve more than 90% accuracy. At present, the AI large model is mainly used in auxiliary work.

In addition, Wang Boyang said that there is still a problem with the accuracy of the application of generated AI in the vertical field. "although generated AI can adapt to all kinds of queries, there may be a lack of consistency in continuous conversations. In addition, generative AI may generate unexpected answers when answering questions.

There are many ways to solve the hallucination problem of generative AI. For example, Xinmei's large model applies self-developed "white box" display, which clearly shows the reasoning process when answering questions, so that the results can be traced.

Tong Guohong explained that insurance products and services usually involve complex risk assessment and decision-making processes, and the application of the "white box" model can better explain its working principle and decision-making basis to regulators, customers and internal teams. it can be more clear and transparent in the consideration of problem tracing and accuracy.

At the same time, the risks in the application of AI technology can not be ignored. Tong Guohong said that the insurance industry involves a lot of sensitive personal information, and how to ensure the privacy and security of the data while using large models is an important challenge. There are a variety of data sources in the insurance industry, how to ensure the quality and standardization of data, so that large models can effectively learn and predict, is a key issue.