Hit by a hurricane? InsurBot can help with your claim

By and Aisha Walcott | 2 minute read | September 19, 2019

hurricane flooding in Germany

Call centers at insurance companies face enormous pressure when natural disasters strike. Policyholders hurting from their losses need urgent help. But customer service agents can be overwhelmed by the spike in calls.

That happened in recent years when a record number of hurricanes, floods and other disasters hit the United States. Policyholders were turned off spending an hour on the phone to file or check on a claim. And they rightly expect fast service during routine calls to find out about their coverage or ask a question about their bill.

When a call center is inefficient, such interactions can cause significant customer pain points. This is leading many insurers to mitigate them through call-center automation.

Our solution at CodeObjects is a virtual assistant powered by IBM Watson AI technology from the cloud. Called InsurBot, it helps insurers deliver a superior policyholder experience. Once we train the virtual assistant to converse like a five-star agent, every single call has the potential to become a five-star conversation.

Powered by natural language processing

We designed InsurBot to automate repeatable tasks that agents perform every day, such as helping policyholders file a first notice of loss, check a claim status, get an insurance quote and make a payment. The virtual agent responds to queries in natural language following scripts developed from Watson’s analysis of 50,000 calls received after Hurricane Irma.

Although we aimed to make the virtual experience easy and seamless for policyholders, we also wanted to give them a sense of control. Users can reach InsurBot through voice, text or browser and it will refer them to a live agent if they prefer.

A real-world test for managing insurance claims

Just as our developers were finishing the initial scripts, Hurricane Michael appeared headed for the Florida Panhandle. Weather models predicted one of the largest storms to hit the area since 1992 and, inevitably, corresponding catastrophic property and casualty losses.

Anticipating a deluge of claims, Security First Insurance, our customer in the Panhandle, decided to roll out InsurBot ahead of the storm. Going into deployment mode with all hands on deck, we tested it, stabilized it and created the production version that we deployed just before Hurricane Michael’s devastating landfall.

InsurBot performed well for Security First Insurance. It handled 30 percent of the calls received after the storm, completing 75 percent of them without referral to a live agent. This reduced costs and delivered a better customer experience: in contrast to average hold times of 8 minutes and as long as 30 minutes after a previous hurricane, InsurBot picked up instantly.

Promising better service with AI

These successes give us confidence in InsurBot’s promise. In fact, we’ve developed a RESTful API specification that allows integration with any insurance system that supports APIs.

We’ve also added big-picture analytics such as charts that show performance of an intent over time. And detailed analytics can help analysts review specific outcomes down to the transcript level to understand the conversations and identify ways to optimize the virtual agent’s behavior.

InsurBot has shown that a virtual agent can address customer service concerns in the insurance industry. Whether you’re the first person or the hundredth to call, your call should be handled immediately, even during a natural disaster. With auto-scaling enabled for all solution components and an assist from Watson technology, CodeObjects’ InsurBot can help insurers realize that goal.