To handle loss claims after natural disasters, insurers turn to AI
"When customers’ lives are uprooted, they need consistency."
In October 2018, Hurricane Michael struck Florida with 150 mph winds and 31-foot waves. Recorded as one of the strongest storms to hit the Florida Panhandle in 100 years, the hurricane ultimately caused $25 billion in damages, according to the National Hurricane Center.
An estimated three-quarters of Bay County, Florida households were impacted by the storm, which completely demolished some cities.
Insurance providers needed to act immediately to help get people back on their feet. Providers also knew that quickly preparing for a sudden influx of catastrophic loss claims could strain resources.
“A single event the size of Hurricane Michael creates a tremendous amount of pressure for insurance companies to maintain a high level of service,” said Anthony Peccerillo, Vice President of Information Technology & Strategy at CodeObjects, a cloud-based property and casualty insurance platform.
“Having to prepare for one or more catastrophes in a short period can push the capabilities of even the best organized insurance providers,” Peccerillo said.
With Hurricane Michael, insurers anticipated a surge in customer service calls but didn’t know whether they could quickly and seamlessly field them all. Third-party call centers are typically outsourced by insurance providers during natural disasters, but these agents usually do not have the knowledge or authority to answer common policy questions. As a result, wait times can be extremely long.
“Imagine the frustration of a customer when, after waiting 30 minutes to finally report their claim, an agent is unable to tell them the deductible amount on their policy,” said Peccerillo. “This creates a very disjointed, confusing and aggravating experience.”
CodeObjects’ virtual assistant, InsurBot.ai, was created using IBM Watson for this exact purpose. The self-service solution allows customers to ask questions, whether through text, chat, or voice interactions. InsurBot.ai has a thorough understanding of the insurance domain and is able to file a first notice of loss, check a payment or claim status, get a quote, make a payment, or answer questions on deductibles or coverage.
It was first used during Hurricane Irma and analyzed approximately 50,000 calls to learn how to enhance the customer experience.
“InsurBot leverages AI and our insurance domain to consistently handle the most common transactions, allowing customer service representatives to invest their valuable time on elevating the level of service offered,” said Peccerillo.
“We saw a real need to turn poor customer service experiences into exceptional ones by leveraging AI.”
Prior to Hurricane Michael, CodeObjects partnered with client Security First Insurance, a Florida homeowner insurance company. Security First wanted to see how Watson would enhance its overall customer experience during disasters, and integrated Insurbot.ai into its business.
As the hurricane was forming in the Gulf of Mexico, CodeObjects and Security First sped up the implementation process. They fast tracked the project to process the first notice of loss services calls prior to the hurricane. With just a month to prepare, “we were able to move from pilot to production in a very short timeframe,” said Peccerillo.
75 percent of the policyholders who engaged with InsurBot.ai after the hurricane had their issue resolved with zero wait times, and without having to escalate to a live agent. That ultimately saved the company $1 per minute during the post-hurricane call spike. When customers’ lives are uprooted by natural disaster, they need consistency, according to Peccerillo.
Thanks to Watson, Security First was able to quickly reach customers, and make the rebuilding process a bit easier.
“Being able to scale and offer a new level of convenience is no longer in the future,” said Peccerillo. “This technology is here, and companies will need to adopt it to survive.”
Peccerillo’s advice to other insurance companies wanting to quickly develop their own AI assistant?
“Start with a manageable project,” he said. “Don’t try to boil the ocean. Using a fairly common and repetitive business scenario, achieve success, continuously measure, and expand aggressively.”