In recent months, sophisticated AI-powered large language models have exploded in popularity. Many companies are now focused on figuring out how they can leverage these programs. Insurance industry leaders are paying close attention. There are countless possible applications and the cost of waiting could be significant.
According to Techopedia1, a large language model is a type of machine learning model capable of performing various natural language processing tasks, such as generating, classifying, and translating text and answering questions. Large language models are also associated with generative AI – tools that create content such as images and text.
ChatGPT is a notable example. Created by OpenAI2, it can communicate conversationally, answer follow-up questions, and challenge incorrect premises. However, it’s not perfect – today, it sometimes provides incorrect or completely fabricated information. Nonetheless, its ability to provide complex responses has impressed people around the world and triggered an arms race to create increasingly advanced large language models.
Another notable large language model is Google’s Bard3. Like ChatGPT, Bard can give impressive conversational answers, but, in its early development, it also provides incorrect information at times.
In LLMS: What to Do Next, Celent4 warns that a competitive gap could develop between insurers who adopt large language models early and those that do not. New AI programs represent a monumental opportunity in terms of maximizing automation, minimizing risks, improving customer engagement, optimizing product development and pricing, and supporting better data-driven decisions. In light of these potential advantages, the risk of doing nothing could include long-term challenges and diminished profitability.
Some insurers are already on board. According to Insurtech Insights5, during Chubb’s Q1 earnings call, CEO Evan Greenberg announced the company intends to adopt AI tools on a larger scale. Chubb has plans to use AI in numerous ways, including claims, marketing, analytics, customer interface, and customer service.
According to Insurance Thought Leadership6, generative AI can automate routine tasks to improve underwriter workflows involving new business pricing, renewals, endorsements, and cancellations. By automating steps in the claims process and streamlining workflows, generative AI can reduce the amount of human work involved. It can also add significant value in new business underwriting with past quote analysis and other risk selection activities, not to mention the benefit it delivers in its ability to analyze massive amounts of data in real time.
According to Insider Intelligence7, the Swiss insurance firm Zurich is experimenting with ways to use AI such as ChatGPT in claims and underwriting. The company wants to use AI to extract information from long claims descriptions and use claims data to improve underwriting. Zurich also wants to see if AI can help write code to use in statistical models.
Large language models are essentially sophisticated chatbots. As such, they can offer 24/7 personalized communication with customers. Insurers can leverage this both for marketing purposes and to engage existing policyholders and claimants.
Gizmodo8 says many companies in a wide range of industries are already using ChatGPT. For example, Bain & Company wants to use AI to enhance marketing, whereas Instacart plans to use ChatGPT to provide customers with answers to shopping-related questions – for example, when asking for a shopping list to go with a recipe.
Property Casualty 3609 says ChatGPT has some applications in cybersecurity. Although it wasn’t designed for code debugging, it spots code errors and vulnerabilities. Plus, it can examine data for patterns and anomalies during cyberattack investigations.
This application is also important in fraud detection. According to the Coalition Against Insurance Fraud10, fraud occurs in approximately 10% of all property and casualty losses. However, fraud is not always reported or detected. In fact, 29% of people who say they were the victim of auto insurance fraud never reported their suspicions.
New AI tools could help. Insurance Thought Leadership11 says generative AI can analyze large amounts of data to detect patterns or anomalies that indicate fraud.
Although insurance companies may be eager to leverage large language models, they need to be aware of the risks. Property Casualty 36012 warns that some bad actors are using ChatGPT to create phishing emails. In addition, companies using the program for legitimate purposes may unintentionally release confidential information.
According to Dark Reading13, Samsung Electronics has already experienced at least three incidents involving sensitive data leaked to ChatGPT. In one, an engineer pasted buggy source code into the program to gain help with fixing the errors. Incidents like this show that companies need to be aware of the implications of providing large language models with data. For insurers with a significant amount of sensitive data, this is a major concern.
Insurers using large language models to provide policy or loss prevention information need to be mindful of the risk of providing incorrect, changing or overly vague information. Generative AI should be used to assist, rather than to replace humans – and humans should always provide oversight and validate accuracy. Large language tools make mistakes, which is why Carrier Management advises insurers to proceed with their eyes wide open.14
Although it’s important to consider the risks, leveraging technology like large language models and digital payments can give insurers a significant competitive edge. At One Inc we provide seamless digital payment experiences that allow insurers to deliver more to their customers. Learn more.
Sources
Tags: AI
The One Inc Content Team strives to provide valuable insights about digital trends and payments innovation for the insurance community.