This code template tells you how to create a platform where insurance agents can handle claims. We use IBM® Watson ™ language processing features to understand, classify, and retrieve data to reduce repetitive tasks. This in turn allows the agent to handle more creative and complex problems, and most customers get answers to their questions faster with the help of a Watson-based virtual assistant.


We create a virtual insurance assistant using Node.js and Watson Assistant. The facilitator uses Watson Discovery to answer questions about the policy. When processing claims, the assistant uses the Watson Natural Language Understanding tool to identify what kind of correction is needed when recommending a mechanic.

The mechanic recommendations have been built with the help of the Watson Knowledge Studio to create custom entities for mechanic surveys and the Watson Natural Language Understanding program to process estimates and make the best recommendations based on the type of correction. Separate teaching program guides you through creating and deploying a custom template in the Watson Natural Language Understanding. By combining our custom entities with the built-in identification of each review opinion (e.g., positive or negative), we are able to categorize mechanics by emotion for each type of correction. When the customer describes the requirement for a virtual assistant, the model implemented determines what type of repair is needed to narrow down the choice of mechanics.

We made answering policy questions part of the Watson Assistant dialog box. In this case, when Watson Assistant detects that your intention is a policy request, it will forward your question to Watson Discovery. Follow the separate instructions for Watson Discovery to understand the insurance contract documents teaching program, which uses Smart Document Understanding to train Watson Discovery to read sections of an insurance document. The documents are then put into the Watson Discovery collection. Watson Assistant will ask Watson Discovery directly and return the answer to your policy question.

A fully functional virtual insurance assistant must first complete the following guides:

The resulting trained Discovery collection will be used for case studies. The model implemented by Watson Knowledge Studio and Watson Natural Language Understanding is used to understand requirements descriptions.

Once you’ve completed this code template, you’ll understand how to:

  • Process complex insurance documents with Watson Discovery to effectively answer customer-specific questions
  • Use Watson Knowledge Studio to create custom models and entities to better understand and categorize mechanical reviews
  • Create a web-based application with a virtual assistant who can answer policy questions and make recommendations based on which mechanics are closely tested and covered in practice.


  1. Insurance documents are uploaded to Watson Discovery and then marked with the Smart Document Understanding tool.
  2. Mechanical review documents are uploaded to the Watson Knowledge Studio and then tagged to create custom entities and relationships.
  3. The user chats through the web application to chat with Watson Assistant.
  4. Watson Assistant answers policy questions using the Watson Discovery survey features.
  5. The assistant recommends a mechanic based on the types of damage to the vehicle and Customer Reviews.


You will find the detailed steps of this model readme file. The steps show how to:

  1. Clone the archive.
  2. Collect the credentials of the mechanic’s recommender.
  3. Collect credentials for policy surveys.
  4. Create a Watson Assistant skill.
  5. Enable the application.
  6. Use the application.

This code template explained how to create a forum where insurance agents can deal with claims. That’s part of it Build a customer service solution help your customers manage their insurance claims and get car service information.



Please enter your comment!
Please enter your name here