IBM Cloud Paks Applied to a Real-Life Scenario:
Cognitive Insurance Claims (Part 1)
By Michele Chilanti and Sundari Voruganti
IBM Cloud Paks are the centerpiece of IBM’s hybrid cloud strategy. They offer a powerhouse of capabilities and tools that work together on an integrated platform to address even the most advanced business challenges. Thanks to the portability of RedHat Openshift, you can deploy your solutions on any cloud platform.
In this two-part blog, we illustrate how we used multiple IBM Cloud Paks to deliver a real-life solution, inspired by a common scenario in the Insurance industry. We discuss the creation, deployment and the operational aspects of your solution using IBM Cloud Paks.
The application we are going to demonstrate here was originally developed by the Insurance Industry Solution team within the IBM Global Industry Solutions organization.
The business scenario
The problem we are solving is centered around the auto insurance claims processing. Drivers – occasionally – have car accidents and need to contact their insurance provider to submit a claim. Providers then need to gather all the necessary information, evaluate the legitimacy of the claim, estimate the extent of the damage, assess the potential for fraud, in addition to keeping their customers happy by helping them with arrangements for rental vehicles, tow trucks, and repair shops.
Traditionally, this is a “high-touch” serial process flow, where claim administrators, adjusters, and back-office analysts take turns at performing all those tasks in a mostly manual fashion, which impacts costs, efficiency, rework, and scalability.
With a combination of AI, analytics, business process automation, and integration technologies the vast majority of those tasks – in fact, all of them in many cases – can be fully automated, thus dramatically improving the timing, consistency, and overall cost of the end-to-end auto claims process.
Let’s explore how the IBM Cloud Paks make this possible.
The diagram below shows the conceptual functional flow of the solution we developed – Cognitive Insurance Claims.
The diagram illustrates the high-level steps of the cognitive claims solution:
- The driver has an accident. Using a mobile app provided by the auto insurer, the driver initiates a dialog with the insurance company (which uses a virtual assistant), and signals the intention to create a claim.
- The assistant gathers all the necessary information: it confirms the identity of the claimant, gathers the identity of the other party involved in the accident (and their insurance information), and asks about time and location of the accident.
- The system creates the claim and stores all the relevant information with it. The claim is now visible to anyone in the back office through a dedicated web portal, as shown below.
- The assistant asks for photos of the cars involved in the accident. It stores those pictures with the claim. Image recognition functions assess the type and entity of the damage. Subsequently, a machine learning model is invoked to assess the likelihood of fraud and the complexity of the claim. If both checks come back negative, the claim can be resolved with no human intervention.
- At this point, the assistant notifies the driver that the claim has been approved and offers to make arrangements for repair shops, rental cars, taxi rides. All this is also automated by the underlying business process.
How we did it
So – how to leverage IBM Cloud Paks to create the flow described above?
This architecture diagram shows you the ingredients we have used for the implementation.
As shown above, we have combined the forces of three IBM Cloud Paks:
- Cloud Pak for Data – which provided the tooling and runtime support for the Machine Learning models that perform fraud detection and complexity analysis. This Cloud Pak also provides Watson Assistant, which we used for the implementation of the virtual assistant (although we used a SaaS version of it, hosted on IBM Cloud).
- Cloud Pak for Business Automation – which provided the tools and runtime for the business process orchestration that guides the overall execution of the claim processing. The capability we leveraged is called Business Automation Workflow.
- Cloud Pak for Integration – which provided the integration bus (App Connect Enterprise) that runs the integration flows that make it possible for the claim information to flow into the insurance company’s back-end system of record.
Note that we have Cloud Pak for Watson AIOps in the picture, which will be discussed in Part 2 of this blog, when we talk about the operational aspects of the solution.
The solution also includes a set of microservices developed as traditional REST API providers – in NodeJS – which act as the back-end glue code for the mobile application. These APIs are also exposed for potential third party consumers (channels and partners) via API Connect, thanks to Cloud Pak for Integration.
The power of IBM Cloud Paks – besides the functional breadth they provide – is also magnified by the fact that all those capabilities are accessible through a common user interface and through common gestures and administrative actions. Here is a snapshot of the Cloud Pak for Automation Dashboard, which shows how you can access both automation and integration capabilities from a single, integrated user interface:
What about operations?
So far, we have touched on the functional aspects of the solution – you might be curious, at this point, as to how one can effectively monitor and manage the operational environment that hosts all these components.
Part 2 of this blog shows how we leveraged IBM Cloud Pak for WAIOps, Instana, and Turbonomic for that purpose.