4 ways NLP technology can be leveraged for insurance

By Katerina Mansour - 18 December 2020

Digitisation in the insurance sector has generally been slower than in other industries. Heavy regulations and complex legacy systems have been the main hurdles to adoption of digital processes.

However, more and more key players in insurance are embracing digital solutions to their pain points. One of the many technologies growing in adoption in insurance is natural language processing (NLP).

What is NLP?

NLP, a subset of artificial intelligence, provides the ability to automatically read, understand and derive meaning from text in a variety of contexts. The technology can be trained to be particularly knowledgeable in specific areas – an essential requirement for medical or insurance use cases.

In insurance, this is typically done by having the technology analyse a large number of claims in order to create a database of knowledge. Startups and large technology providers leverage NLP to provide insurers with tools to speed up decision-making, reduce costs and avoid human errors.

The global NLP market is expected to be worth $35 billion by 2025. In a 2019 survey, 84% of insurers said they believed AI will revolutionise their sector. Furthermore, Accenture’s Living Business survey found that 64% of the time, when a customer changes from one insurer to another, it’s to seek a more relevant product, service or experience. As such, embracing AI solutions is also relevant to provide customers with the experience they’re now expecting.

So, how exactly is NLP being leveraged by insurers today?

Source: Bold360.

Customer service and assistance

According to an recent survey, more than 80% of insurance customers want personalised offers, messages, pricing and recommendations from their auto, home and life insurance providers. Virtual assistants, powered by NLP technologies, can help deliver that personalised experience to insurance customers.

Beyond automating customer experience personalisation, is the need to provide them with timely and accurate answers to their questions. An insurance agent arguably wastes a considerable amount of time on low added-value tasks like answering basic questions.

IBM has estimated that within a 6-minute customer service call, agents spend 75% of that time doing manual research to find the relevant information. Thus, insurers can use NLP-based solutions to speed up common customer service processes.

  • Earley Information Science and Allstate partnered to develop a virtual assistant called ABIe. The technology was developed to help the insurer’s agents better know their products, when Allstate launched its new business insurance division. ABIe has the ability to process 25k inquiries per month, leveraging NLP. It helped make the corporate’s agents more self-sufficient and better sell products to their customers.
  • Accenture has developed its own NLP-based solution, MALTA, that automates the analysis and classification of textual information to help insurers easily access key information. Accenture claims the solution provides 30% more accurate classification than when the process is done manually.

Claims management

Claims processing is a core activity within the insurance sector that is at the centre of many pain points. The task can be time-consuming, costly and subject to human errors. Insurance agents can use NLP during phone calls, for instance, to recognise a client’s speech and automatically fill out a claims form. Overall, NLP technology analyses both speech and text faster than humans can. Employees then simply require to manually verify the results.

  • Lemonade has made its mark by providing personalised insurance policies and quotes to customers through the use of its chatbot, Maya. One of its success stories was its chatbot’s ability to process a theft claim for a $979 lost coat within 3 seconds. This process included reviewing the claim, cross-referencing it with the customer’s policy, running 18 anti-fraud algorithms, approving the claim, wiring instructions to the bank, updating the customer and closing the claim.
  • Sprout.AI uses NLP and optical character recognition to analyse and understand unstructured data from insurance claims. It then pairs this data with external real-time data (weather, geolocation, etc.). The startup’s technology can settle claims within minutes, while also checking for fraud.
  • Similarly, IPsoft develops Amelia, a virtual assistant that can process claims by directly conversing with customers. For example, if a customer needs to file a claim after having a car accident, Amelia will pull up their policy and data, confirm their identity and walk them through the process of filing their claim step-by-step.

Fraud detection

A large variety of traditional and experimental technologies have been leveraged in combatting insurance fraud. NLP technologies are among the most promising options according to experts in the field. Manually reviewing claims notes and data like emails or text messages can be incredibly time-consuming. NLP solutions are able to automatically analyse and understand all this unstructured data (messages, social media posts, claims, etc.). These solutions can then flag cases of suspected fraud for human review.

  • Shift Technology, ranked in the top 5% of Early Metrics’ rated startups, develops technology to help insurers detect fraudulent claims. Its software, FORCE, applies a variety of AI technologies, including NLP, to assign each claim a numerical score representing the likelihood of fraud. The startup recently signed a partnership with Central Insurance Companies to detect suspicious behaviours in its auto and property claims.

Underwriting automation

Underwriters have to analyse a large number of policies and documents to make key decisions. The outcome of their decisions depends on how well and how accurately they have analysed the necessary information. This part of the insurance process is highly error prone by nature. NLP solutions for underwriting use cases are able to extract relevant information to help underwriters assess risk levels.

By extracting key information like dates, locations, names, diagnoses, lab results, etc, these solutions help underwriters access information they would have taken hours to find by parsing through documentation themselves.

  • DigitalOwl and Zelros are two startups that develop solutions to analyse, understand and extract relevant information from documents to help underwriters make their decisions faster, with more accuracy.
  • A Cognizant case study described the ability to predict flood risks in the United States in order to better underwrite policies. The tool helped its client, a global reinsurer, expand the range of data that informed its decision and reduce manual efforts by its underwriters. In the end, it could more accurately define risks, refine its policies and improve case acceptance rates by 25%.

Where does NLP bring the biggest benefit in insurance?

According to a Celent report from April 2020, 67 percent of large P&C insurers indicated that cost reduction and process improvement has become more important due to Covid-19. The insurance sector is one of many industries that are subject to financial and logistical challenges due to the pandemic. NLP technologies can provide solutions that will help insurance players reduce costs, save time and adapt to this new environment.

A 2019 study by LexisNexis found that 88% of insurers surveyed were already seeing benefits from the onboarding of AI-based solutions for claims settlements. The results were promising all around. Overall, it seems the insurance sector might be catching up in its adoption of new technologies. However, key challenges remain, such as access to employees with the necessary expertise, or the lack of quality data available to train datasets.

All articles