ben townsend

In the last few years, companies within the insurance industry have been looking to the growing field of Insurtech to remain competitive and better serve customers. Much of the interest is focused on the ways in which Artificial Intelligence can foster greater efficiency in how those companies do business. A recent survey of senior insurance executives found that 87% are investing $5 million or more in AI annually, with many seeing the potential for that number to increase in coming years. But amid the rush to implement new machine learning systems in an industry that has traditionally been hesitant to embrace technology, it’s worth taking a moment to examine the best ways that AI products can be applied to different aspects of the insurance business.  

The most beneficial thing an AI product can accomplish for an insurance company is freeing up time for employees so that they can focus on more important tasks, and the technology that is most suited to that goal is Natural Language Processing (NLP). An NLP program is one that’s designed to read and understand text at a level of complexity approaching that of a human, but with the speed we’ve come to expect from anything with a microprocessor. NLP is not necessarily the type of technology that most people would immediately associate with AI, which is often identified with the creation of complex risk models or analysis of structured numerical data. NLP instead helps insurance companies create efficiencies for employees who have to spend their days dealing with large amounts of human-generated text. 

NLP has potential applications for the insurance industry in both customer service and internal operations. On the customer-facing side, it’s mainly used to power “chatbots” that interact with customers via online messaging to fulfill basic requests and collect information. It’s a useful technology that’s being adopted in almost every industry, and insurance is no exception. The basic goal of the NLP-powered chatbot is to eliminate the need for a customer service worker to deal with the most straightforward questions that come up over and over. Those questions are instead handled by the computer, so that employees may focus on higher level and more complex customer problems. 

But what about the workers outside of direct customer service? Especially the ones who, rather than having to field questions from the public, instead must go through massive troves of written documents. That’s where we can start to see the potential for NLP technology to assist the work  of employees involved with claims, underwriting, payment processing and beyond.

Take for example the work of a claims adjuster at a medical insurance company. It is a basic part of their job that they must spend hours upon hours reading through doctor’s opinions, depositions, and claims files. Imagine if that adjustor had access to an NLP program with the potential to quickly process those tens of thousands of words of text, all written by different authors with distinctive professional contexts and writing styles. The NLP could produce a succinct summary of the documents with references to relevant portions in order to double check, if necessary, the NLP’s report. This would significantly cut down on the time spent staring at PDFs on a computer screen or printouts. The most complex claims usually tend to be the most expensive. The more time an employee with specialized knowledge can devote to high-level tasks and decision making, rather than pouring over text that must be reviewed but might not ultimately prove useful, the more effective that employee can be. 

The most important work at any insurance company involves identifying patterns. The “Big Data” AI technology focused on numerical analysis, as discussed earlier, looks for patterns in the numbers that are fed into the program. NLP works in a similar way, but the goal is instead to  detect patterns in unstructured text data. When looking at a high value claim, for example one filed by someone who has suffered a catastrophic injury that prevents a return to work, an adjuster might use a simple keyword search of the documents related to a claimant’s health and employment. Finding certain words or phrases can make the work go faster and potentially avoid hours of reading, but it has its limits. NLP technology functions according to a similar concept, but on a much more advanced level. Because NLP can understand the nuances of human language, it can find patterns that are deeper and more meaningful than simple keywords and repeated phrases. The technology can ultimately provide an incredibly significant compliment to the knowledge of a highly specialized expert. 

The reason that the insurance industry has such an acute interest in Insurtech and AI is because they have so much potential to transform the industry. No company wants to be left behind and lose a competitive technological advantage. It’s important for decision makers who are exploring ways to incorporate AI technologies into their businesses to look beyond the most obvious applications. Because of the insurance industry’s need to deal with large amounts of unstructured text and identify patterns within them, it has the opportunity to pioneer NLP uses and lead the way in the adoption of AI by knowledge workers.