Visual AI: Second sight for insurers 

Insurance has always been a labour-intensive profession where detail and accuracy are vital. Assessing risks, calculating premiums and reviewing claims are time-consuming and demanding activities. Insurers who can increase precision here while saving time will therefore gain a strong competitive advantage. This is where automation and artificial intelligence (AI) can generate big gains in terms of straight-through processing and customer experience.

Most people are familiar with AI-driven chatbots, typically used to guide customers on insurance journeys, but there are many other ways that AI can add value at key points in the insurance chain, from the application stage through to underwriting and claims handling. One example is the innovative use of visual AI, or computer vision, to support damage reports in car and property claims.

Finding the focus

Insurers are keen to harness advances in AI technologies such as machine learning and deep learning which mirror the way humans build knowledge through experience and neural pathways. With visual AI, the tipping point came after academic research in the UK inspired the launch of Tractable, which enables insurers to harness the power of computer vision to assess customers’ cars or homes. 

“AI began to be visually trainable around 2014,” says Jimmy Spears, Head of Automotive at Tractable. “By that I mean it acquired the capability of the human eye and could make fine distinctions between shapes and objects. Automotive accidents are an obvious use case. There are millions of classified images of car damage, from fender benders through to total losses, and AI can be trained to identify the degree of damage from these images. What you have is a huge digital data trail, an AI gold mine to enable accurate reports and damage estimates.”

Spears says that the accuracy of visual AI is equal to if not better than a human appraiser and provides far superior speed and customer focus than manual processing. “AI eliminates the friction of in-person appraisals,” says Spears, “and puts a visual expert at the heart of your workflow and in your customers’ pockets.”

Visual AI in action

The reference to “customers’ pockets” hints at how the process begins. Following a car accident, customers will take an image of the damage using a smartphone camera. “Connected technology sets the ball rolling,” says Spears, “and there’s no need to download an app, as your insurer provides a weblink for you to send an image from your phone. In addition to analysing static images, Tractable’s technology can also collect live data from a video of car damage and generate on-the-spot feedback.”

After photos are delivered at first notice of loss (FNOL), AI analyses the information and assesses the damage without delay. This is where deep learning through visual training comes into play. “The AI performs triage to accurately calculate repair costs,” says Spears, “resulting in much faster response times. Without AI, assessments can take days or weeks, but with AI you can complete an appraisal from FNOL photos in as little as three minutes.”

Spears says that the increased speed and efficiency help everyone in the insurance cycle: claims teams, repair specialists, third parties involved in claims, and, of course, policyholders. Estimates are provided instantly, parts can be procured swiftly and delivered to body shops, and repairs can be completed without hold-ups.

Targeting the property sector

Visual AI also has enormous scope for property insurers. Having proved the benefits for car insurers and their customers, Tractable is now applying the same technology to assess property damage, which Spears says is a growing challenge because of the rise in catastrophe events.

“Climate change is leading to more frequent and severe natural disasters,” says Spears, “but AI can greatly improve disaster recovery. As with car assessments, AI will generate accurate reports based on photos or videos sent via smartphones. This eliminates the need for an appraiser to view the property in person, and it’s easier to prioritise urgent cases.”

Spears highlights the example of typhoons in Japan, where thousands of people suffered property damage in 2021. “It’s difficult for insurers to manage the spike in home insurance claims after disasters on this scale,” he says. “And if you’re dependent on a human workforce, people can wait months for on-site appraisals before receiving pay-outs. Tractable’s AI property solution, deployed for the first time in Japan, reduced cycle times from claim to settlement to as little as a single day.”

Vision of the future

Versatility is a key advantage of visual AI. Spears says it can be transformative for any insurance activity that requires visual assessments. Other industries and sectors, such as agriculture and manufacturing, can also benefit. “Car and property claims are the immediate and obvious use cases,” says Spears, “but any image or video has a story to tell. It’s the job of visual AI to tell that story completely, accurately, and much faster than a human can.”