Digital Developer Conference: Hybrid Cloud. On Sep 22 & 24, start your journey to OpenShift certification. Free registration

Build a customer care solution

The concept of self-service help is a common trend in many industries, not only due to the current health crisis but also due to companies needing to add efficiencies to their business model to focus limited resources on more strategic and complex initiatives. Enterprises are taking advantage of the rapidly growing availability and efficiency of artificial intelligence (AI) services to automate time-consuming workflows and quickly assist their customers.

This solution demonstrates how to build a chatbot that answers insurance policy questions, helps customers submit claims, and even recommends mechanics based on the type of repair needed. The following demo shows how a customer might interact with the virtual assistant.

via_gif

In this solution, an auto insurer uses AI services such as IBM® Watson™ Discovery, Watson Knowledge Studio, Watson Natural Language Understanding, and Watson Assistant to help customers manage their insurance claims and get automobile service information in a timely fashion. Automobile policies can be challenging to customers because they can be complex and lengthy. Additionally, many people can relate to needing to find a reputable mechanic. AI solutions can be trained to aggregate, filter, and quickly provide the information that is needed when it is needed.

By combining the Watson technologies, you can:

  • Help customers easily filter and view policy content relevant to their current needs
  • Help customers quickly identify who is the best candidate to service their vehicle
  • Identify automobile service trends in a region to help in setting or adjusting rates
  • Identify customer needs and proactively provide follow-up and additional services

This solution uses Watson Discovery to process unstructured policy documents to answer customer questions and Watson Knowledge Studio to create custom models that classify mechanic reviews so that insurance claims can be matched to the appropriate type of repair shop. It uses Watson Natural Language Understanding for consumer sentiment analysis to rank the mechanics.

These Watson services are then integrated with Watson Assistant to create a self-help customer service chatbot that can be added to a company’s web site. Customers can interact with the chatbot to get the help they need without waiting for a human agent.

Architecture flow

Customer case solution flow

  1. The developer trains Watson Discovery to recognize policy documents.
  2. The developer collects and annotates domain language examples and publishes to the Watson Natural Language Understanding service.
  3. The customer loads the web application and interacts with Watson Assistant.
  4. Watson Assistant forwards policy questions to be analyzed by Watson Discovery.
  5. Watson Assistant forwards claim descriptions to be analyzed by the custom Watson Natural Language Understanding model.

Although the theme for this content is around an insurance industry, the framework can be applied to any enterprise wanting to provide a self-service option to their customers, such as health and home insurance.

This solution contains three parts:

  • Process, understand, and answer policy questions with Smart Document Understanding
  • Build a recommendation engine with Watson Natural Language Understanding
  • Build a virtual insurance assistant to help process claims

Process, understand, and answer policy questions with Smart Document Understanding

Watson Discovery uses AI search technology to retrieve answers to questions. It contains language processing capabilities and can train on both structured and unstructured data. The data that Watson Discovery is trained on is contained within a collection (that is, a database).

As a first step, the solution ingests a sample insurance document and uses the built-in Smart Document Understanding annotation tool to train Watson on the different sections of the insurance document. This way, you can break up the document, omit certain sections, and improve the query accuracy.

You use the Smart Document Understanding tool to extract custom fields in your documents, allowing you to customize how your documents are indexed into Watson Discovery, which improves the answers that your application returns.

In the Process, understand, and answer policy questions with Smart Document Understanding tutorial, learn how to:

  • Acquire sample insurance documents
  • Create a Watson Discovery collection by uploading an insurance document
  • Configure, annotate, and filter the data by using Smart Document Understanding to improve the accuracy of responses to Watson Assistant, break up the document contents, filter out unnecessary sections, and annotate it
  • Train Watson Discovery on insurance-domain terms
  • Test the model within Watson Discovery by asking basic insurance questions in natural language

Using SDU

Build a recommendation engine with Watson Natural Language Understanding

Watson Knowledge Studio allows you to teach Watson domain-specific language, with custom models that identify entities and relationships that are unique to a specific industry in unstructured text. Watson Natural Language Understanding uses deep learning to extract metadata from text such as entities, keywords, and categories.

In the Build a recommendation engine with Watson Natural Language Understanding tutorial, you extend the first tutorial by:

  • Collecting and uploading text documents that describe automobile damage and repairs
  • Using Watson Knowledge Studio to annotate and classify reviews for auto repair shops
  • Training a machine learning model that is able to determine which are the best shops for specific automobile repairs
  • Testing the model by deploying it to a Watson Natural Language Understanding cloud instance

relations.png

Build a virtual insurance assistant to help process claims

To complete the solution, you integrate the Watson services with Watson Assistant to create a self-help customer service chatbot.

Watson Assistant uses Watson Discovery querying capabilities to help answer policy questions for the customer. The chatbot helps process claims and responds with mechanic recommendations using Watson Natural Language Understanding with a custom domain language model.

In the Build a virtual insurance assistant to help process claims code pattern, you complete the solution by:

  • Creating your dialog by importing your Watson Assistant dialog skill
  • Integrating your Watson Discovery instance with your virtual insurance assistant so that your assistant can search for answers through the Watson Discovery-trained model
  • Deploying an application and testing out your chatbot

Virtual insurance assistant