“..It’s usually a red flag when someone says we want to use Machine Learning to solve this task I’m like that sounds like bull sh*t…” – Elon Musk, Aug 2021 during press release of Project Dojo.
Pandemic accelerated digital transformation across businesses. Overnight workforce moved remotely. Business and Consumer facing organization have rapidly digitized workflows and digital customer touchpoints have surged[i]. While organizations did not invest directly in Artificial intelligence, growing digitization means businesses have more access to customer data which was previously not available.
Surge of AI platforms like TensorFlow & Sage Maker are enabling businesses to cut down cycle time to roll out Machine Learning products. But using AI to solve business problems is not an easy thing to do! Firstly, a bulk of Machine Learning involves getting data ready but more importantly organizations must identify a business case where AI can help organizations achieve desired outcomes. Here are some successful applications in commercial property insurance:
Homeowners in areas of wildfire risk are not only threatened by direct effects of property and casualty loss but also risk decreased coverage and surge in premium in case of an event. Zesty.ai uses data from hundreds of properties across various wildfire events along with artificial intelligence to assess the impact of wildfires on real estate and personal property. Zesty.ai enables homeowners to reduce probability of loss from wildfires. In a matter of few clicks using their smartphone camera, homeowners can assess risk to their property. Customers get a list of recommendations from managing brushes around the house, to using fire retardant construction materials that will minimize threat from wildfire.
Flyreel guides homeowners and commercial property owners with interior inspections using their smart phone cameras. In post-Covid world homeowners are averse to letting outsiders inside their property for inspections. Flyreel app found its timely application in that it enables property insurers to better understand the property they are trying to cover without physically entering the property, assess the risk and offer competitive rates. By off-loading the inspection process through self-service to the insured, Insurers no longer must calculate based on risk pools i.e., exposure based on prior loss in the area.
Underwriters continue to be an integral part of underwriting process. Their judgement is key, since if it were not for Underwriter judgement the whole process could be completely automated. While underwriting continues to work within the constraints of modeled premium and state regulatory boundaries, the nature of data feeding into underwriting rules frame work has become real-time in addition to traditional historic data. Planckdata collects data from posted photos, public data and even customer reviews to synthesize risk signature of commercial business. The availability of new data means Underwriters will need to excise more judgement on top of the rules driven actuarial models.
In addition to changing nature of technology and data capture, the nature of risk itself is changing. Insurers are challenged with offering competitive products to looming and previously unseen risks. With increasing shift to working remotely, the boundary between work and personal use of devices and vehicles are blurring as well. So, while adoption of technology and data is important for underwriters to stay relevant, insurers will need to create right organization culture that fosters innovation and flexibility.