By: Amit Unde, Vice President & CTO, Insurance, LTIMindtree

Artificial Intelligence (AI) has emerged as the most promising technology over the last couple of years. With easier access to technology capabilities offered as a service by multiple vendors, businesses can experiment at much lower costs and get significant benefits, not only to achieve radical cost efficiencies, but also to acquire a competitive edge and differentiate.

This technology is more relevant for the insurance industry, as the fundamental concept of insurance is based on finding patterns in historical data and predicting an outcome. Most insurers in the industry have been leveraging advanced analytics on a structured data. With advent of AI, it is now possible to analyze unstructured data such as images and videos. The machine can now learn patterns based on a historical data and predict an outcome, and continuously fine-tune to achieve better predictability. One of the interesting capabilities of AI is to process natural language to identify the intent, and carry out a human like interactions.

These capabilities have lent themselves to solve many of the insurance problems, and many use cases have emerged. The use cases can be divided into following categories –

1. Cognitive Edge – Ability to do what was previously not possible with traditional technologies, such as image/video analytics and finding patterns in unstructured data.
2. Cognitive Engagement – Ability to engage with customers and interact.
3. Cognitive Process Automation – Ability to automate business processes that are contextual and not strictly rules-driven.
4. Cognitive Insights – Ability to apply machine learning to discover patterns and insights within datasets.

Insurance Use Cases:

Cognitive Edge: Policy in Minutes

With image analytics, it is possible to analyze a selfie and determine multiple parameters, such as the person’s age, BMI, habits (e.g. Smoking), etc. that are important from life insurance underwriting perspective. This information can help to determine if a medical underwriting is required or not. If it is not required, it is possible to provide an instant quote and bind a policy within minutes, as against days that are required today.

Cognitive Edge: Zero Touch Claims

In P&C insurance, the image analytics technology can be applied to analyze the images of cars in accident, and determine parameters, such as make and model of the car, parts that are damaged, and assess the replacement costs. For a small, fender bender accidents, it is possible to adjust the entire claim in minutes without any human interaction.

Cognitive Engagement: Virtual Agents for Customer Service

The natural language processing capabilities have enabled insurance companies to develop Chatbots that are available 24X7, and can answer most of the customer service requests and questions. The Chatbots are savvy enough to understand what they can handle and transfer over to human agents if the requests are not in their domain. This has a huge potential to improve customer service, especially with millennials as ‘texting’ is a natural interface for them to seek information.

Cognitive Process Automation: Cognitive Data Intake

Given the less adoption of standards in the industry and proliferation of unstructured documents, there is a high potential of applying AI to reduce inefficiencies in the insurance processes. One of the key problems in the industry is around data intake from business partners. For example, when insurance carriers get submission data from brokers, it comes in variety of formats without any standardization. An army of people is required to map the data to standard format and process the submission. The mapping is not strictly rule-based, however, there is a pattern in mapping that machine can learn and automatically detect the mapping for new submissions. It can also be used to improve data quality and enrichment by detecting gaps in incoming data and invoking services to curate it.

Cognitive Insights: Machine learning in Underwriting

Many carriers have predictive models to determine the maximum possible loss, probability, pricing and RoE, etc. that are critical to run the insurance business. However, as the customer demography, risk characteristics, market environments continue to change at a rapid pace, these models get out of tune with reality, and it is difficult to keep up with the changes. With AI, it is possible to provide a feedback loop for machines to learn and adapt to changing business needs.

Also, the underwriters deal with multiple unstructured documents, such as Underwriting Guidelines, Reinsurance contracts, Loss sensitive pricing, etc. Each of this document has multiple pages, variety of structures as per business partner formats. It is a tedious and error-prone job to go through documents and extract information for making business decisions. The AI engines can extract the information from unstructured documents and align it to common vocabulary, and make the information easily accessible through a search engine or virtual assistants.

Implementing AI for Insurance

Although there is a large potential of AI in insurance, there are many obstacles in taking the most advantage of the technology:

• Lack of good data sets
• Overwhelming efforts for preparing the data for processing
• Lack of understanding of domain
• Evolving technology with variety of options
• Lack of standardization with too many point solutions

LTI’s Mosaic Decisions and AI platform offers such a framework to insurance companies. Some of the key features are –

• Workbench studio for preparing training datasets
• Configurable data flow with ease of consuming popular A.I algorithms and services
• Multiple channel adaptors – Mobile Apps, Facebook, Skype etc
• Ready insurance Vocabulary, Metadata
• BOT Engine
• Enterprise Adaptors with insurance systems (such as Policy administration, Claims, CRM and Billing systems)