Next-gen data powers insurance risk assessment and response in warming world

According to the World Property and Casualty Report, artificial intelligence- and machine learning-based pricing models, leveraging increased computing power in the cloud, can be used as vital tools in data analytics, weather prediction and loss assessment. Through cloud technology, insurers are able to manage, process and retrieve vast amounts of data – data that can be used in finely tuned AI and ML-based pricing models. Creating such models allows insurers to accurately predict the exact locations of potential climate-related risks and provide customers with actionable steps to take prior to, during or after a natural catastrophe.

“Many innovations hold promise for reducing climate risk. Virtual underwriting and claims solutions, supported by improved digital camera and video capabilities, can introduce efficiencies that help remove vehicles from the road,” writes Ryan Vigus, Executive Vice President at CSAA Insurance Group, in an email interview with Digital Insurance. “In addition, AI-driven risk management platforms apply machine vision and geospatial imaging to anticipate, reduce and manage risk. We monitor, invest in and implement technology solutions that we believe have the potential to help our customers understand climate risk and take actions to protect themselves from the challenges introduced by climate change.”

Digital also provides the opportunity to offer insurance and loss or claims assessments at a larger scale, in cases of major natural disaster. Raincoat, a Puerto Rican insurtech, offers parametric insurance – so named because it pays instantly once a certain parameter has been reached, such as wind speed –  through digitally based solutions that process large-scale individual claims in response to devastating hurricanes, floods and earthquakes. 

The company was born from the aftermath of the 2017, category-five Hurricane Maria that devastated Puerto Rico – millions were left without electricity, water or aid for months, even years. Jonathan González, CEO and Raincoat founder, returned to Puerto Rico to help his mother, who was one of many impacted by the storm. After waiting nearly a year before a claims adjuster appeared to assess the damage of her home, González knew that there had to be a faster way for affected individuals to receive aid. 

“I started researching with my team and what we found was that there was this concept called ‘parametric insurance,'” González explains. “But what we were confused about was that we had never seen an application at a consumer level.”

After meetings with large insurance companies, government officials and regulators regarding the subject, González realized that the difficulty stemmed from the lack of an appropriate risk model for such an event. 

“You need to figure out how you’re going to detect events in real time. You need to integrate that all into some data platform. So all of a sudden, it stopped being a financial issue and started being a software issue… And that’s when it clicked for us,” González says. “We can consolidate that software into a unified platform… and provide that to insurance companies or brokers or governments as a way for them to build and execute their own solutions.” 

Raincoat’s modeling solutions first provide an assessment of the intensity of the natural disaster, and then determine the pricing of the loss payouts based on the calculated damage to an area. Their wind field modeling system specifically estimates the impact of winds in a community, and thus, how the storm will impact an area. Raincoat’s digital platform also alerts policyholders of all transactions in real-time with individual policy tracking and transparent claims payout alerts.

Another insurtech embracing the power of AI technology is Zesty.ai, a property risk analytics platform that offers insurers and real estate customers precise information on over 200 billion data points, including property values or characteristics, as well as potential risks of natural disasters such as wildfires, severe convective storms and wind or flood disasters. 

Partnered with national and regional insurance companies, reinsurers and MGAs, Zesty.ai uses its technology to understand key property characteristics and to build predictive climate risk models. The company leverages AI technology, such as computer vision, with aerial imagery, satellite imagery, building permits, weather station data, transaction history and data from sensors to process risk factors and characteristics of each property. 

“Our focus is very much property specific,” explains Attila Toth, Zesty.ai CEO and founder. “How complex is that roof? What is the condition of that roof? How much of the roof line is covered by dry brush? Those are some of the questions that our computer models answer.”

Zesty.ai’s “Z-FIRE” risk model uses aerial imagery pre and post-wildfire events to process properties that have been fully destroyed, partially affected or left intact in a wildfire event. Z-FIRE aerial imagery, which can be updated daily, can even identify significant individual property characteristics, such as vegetation density surrounding the property and how that may be affected.

“Our approach is rooted in science, but it’s also rooted in understanding what happened in the past. The past is not always a 100% predictor of the future, but if you can build the largest loss database at the individual property level from 1,500 individual wildfires dating back about a hundred years in California, about 20 years outside of California, then you have a better chance of modeling the future,” says Toth. 

At Verisk’s climate panel, Carlos Martins, Senior Vice President of Verisk Claims Solutions, also discusses the use of aerial imagery in the insurance industry and how such technology can be used in the aftermath of a natural disaster – especially to combat cases of fraudulent activity. 

“Insurers have never been positioned better, thanks to the advancements of technology to respond to a catastrophic event,” Martins says. “Thankfully many of the insurers have invested in the technologies and the tools, the analytics to detect this type of activity and intercede and know when they need to send out a special investigator, as an example, to investigate a loss.”

According to Martins, insurers have access to two key types of aerial imagery:  “Blue sky imagery,” which provides data on properties prior to a catastrophic event, and “gray sky imagery,” captured by drones and airplanes so that insurers can begin assessing damage to a property within 24 hours of an event. Through photos and videos, adjusters can capture measurements and dimensions, assess materials and evaluate damage within hours of an event – compared to the traditional claims process, which can take much, much longer. This is an especially significant tool for detecting fraud.

Martins also notes that image metadata, or information embedded into an image, can be used to alert insurers of “a photo that’s being submitted by an unscrupulous contractor, or perhaps an opportunistic policyholder, [that] is illuminated to show inappropriate activity,” Martins explains. “[Insurers] have never been armed with the technology and the tools to quickly respond to a catastrophic event then they will for this upcoming season.”