AI in Insurance: Is It Worth It?
In this article, we explore how AI helps insurance companies to streamline claims processing, enhance customer profiling, and increase fraud protection.
Retooling Claims Processing
With the rapid advancement of IoT and cloud applications, we now have access to lavish amounts of data about the physical world. Countless wearable devices and location-based sensors can tell so much more about people than traditional client assessment models used by insurance companies. With AI, each customer journey becomes tenfold more personalized, services become carefully tailored, and insurance rates more accurate.
Parsyl Insurance shows what the powerful symbiosis of IoT and AI can do in the insurance context. Parsyl’s advanced risk management platform uses smart sensors and ML-based software to allow shippers of sensitive goods to better understand and mitigate supply chain risks. The company’s other initiative called ColdCover leverages IoT sensors installed on carriers’ side to provide insurers with valuable data about the cargo condition. This works both ways, as now customers can make use of cargo settlement predictions and reduce claims cost.
Neos Ventures, a UK-based insurance company, uses IoT and AI to proactively monitor and protect its customers’ homes with the help of smart sensors. The Neos app alerts users about common problems like water or gas leaks and automatically submits claims.
Together with IoT, AI will also play an important role in policy pricing. Traditionally, the insurance company looks through applicants’ driving history, their age and other general characteristics manually. With AI, the insurance company can individually assess the risk involved in insuring a particular person by evaluating their driving style and habits.
This way one of the largest US vehicle insurers Progressive introduced a product called Snapshot, which utilizes the usage-based insurance (UBI) method. Snapshot can gather data via a mobile phone or a special device installed in a vehicle. How often the customer drives and for how long? Does the driver have the tendency to accelerate frequently? How sharp are his or her turns? Answers to those questions are then fed into a predictive analytics algorithm. After six months of analyzing the customer’s driving behavior, Progressive’s employees receive recommendations on the customers’ premium payments adjustments.
The principle behind Neos and Progressive is that they offer lower insurance premiums for those who are willing to install sensors in their homes or cars respectively. Such privacy trade-offs are essential for this AI-based insurance model to operate, and it’s finding more recognition regardless. By 2027, the global UBI market is set to reach $125 Billion, according to Global Market Insights.
While getting a life insurance comes down to a personal choice, car insurance is a liability in many countries in the world. Traditionally, this insurance sector had to rely on repair claims, which are very time- and cost-inefficient. Personal investigation was the only way to make damage assessments. With AI, many integral insurance processes can be automated, saving costs and increasing customer satisfaction.
For example, UK-based insurance company Ageas uses computer vision and machine learning to evaluate accident damage in minutes. After the driver submits photos to the platform, the AI algorithm behind Ageas, developed by Tractable, identifies the affected parts of the vehicle and provides the insurer with a detailed estimation report. Moreover, Ageas recommends repair and paint and estimates how much time the vehicle recovery will take.
Tweet
Iflexion is up to the job.
Customer Profiling
AI can increase the appeal of insurance for a broader audience by utilizing real-time actionable data in place of lifeless, possibly erroneous statistics. Through applying these advanced ways of data collection, customer profiling capabilities significantly increase.
By utilizing facial recognition software, insurance companies can draw accurate insights about a person’s mass index, biological age, habits, life expectancy, and more. This is what made Lapetus Solutions appear in the headlines three years ago with their Chronos technology. The applicant uploads a selfie, answers nine questions, and receives a personalized life insurance policy without the need for a medical examination. For example, AI-based tool can accurately identify if an applicant is a smoker. Interestingly enough, this insurance software takes into consideration even the smallest details like crow’s feet around the eyes.
Equipped with the ability to source both internal and external customer data, insurers can now build a much more detailed picture of their customers. What are their real insurance needs and reasons for application? What are their interests and fears? By deriving these insights from vast data pools, insurers can segment their customers into clusters and recommend relevant product packages for each of them.
For example, US-based Layr, a commercial insurance platform for small businesses, uses machine learning to compare an insurance applicant to clusters of businesses in the same niche. The AI algorithm behind Layr platform automates the tasks of brokers and underwriters, which, in turn, allows business owners to find the best insurance for their needs.
As time goes by, the datasets become bigger, making these systems more precise and reliable. In the long term, AI in insurance will take on the role of a definitive efficacy factor.
Tweet
Reacting to Сlimate Сhange
The effects of a climate change become severe globally and impact many industries including insurance. The year of 2019 was illustrative in this regard. Australia wildfires killed or displaced about three billion animals and caused around 100 billion US dollars in damages to infrastructure. Unfortunately, experts predict that this trend is only going to evolve.
Insurers’ solution to this emerging problem shouldn’t be surprising: they simply increase rates. This makes insurance unaffordable for low-income customers. This, in turn, makes the gap between the total value of damaged assets and the total value of assets covered by insurance wider. This widening protection gap is an emerging problem for our society as a whole.
Logically, the only economically reasonable solution here is to have accurate climate change predictions at hand. Fortunately, modern AI-based predictive models have all the potential to do so. For example, Japan is already using complex computer vision systems and satellite images to predict natural disasters. Our environment incorporates significantly larger datasets compared to businesses, which makes predictions much more accurate. On a societal level, such technological advancements are equal to saving lives and mitigating global risks.
Climate predictions can significantly impact both sides of the insurance industry. Potential victims of natural disasters can now be notified prior to incident occurrence, raising the chances of mitigating the majority of negative consequences. This is what US-based insurance firm Hippo does in practice. The company locates its special IoT devices at clients’ homes and warns them about incoming environmental threats. Moreover, after an adverse event, Hippo’s AI-based platform enables its customers to file claims faster, and consequentially, recover more quickly.
Understory, a US-based tech startup, helps automotive dealers to protect their vehicles from hail storms. Understory’s flagship product Hail Safe makes the notoriously lengthy and complex insurance process smooth and fast. Understory installs smart sensors at a dealership’s location and thus collects accurate weather measurements. When hail reaches specific conditions stated in the contract between the dealership and Understory, the sensor activates claims to the insured. This completely eliminates paperwork and makes insurance much more streamlined.
Tweet
Protecting against Insurance Fraud
The demand for more robust cybersecurity has been constantly rising for the past decade. In the US alone, fraudulent insurance claims accounted for over $34 billion in 2019, according to FRISS. Being able to distinguish between real and fraudulent claims directly corresponds to financial wellbeing of an insurance company. Current systems for detecting insurance fraud are also computerized, but utilize a rule-based approach. In other words, they search for fraudulent indicators and known templates to flag claim as suspicious.
While this system proved to be effective to some degree, wrongdoers never stop coming up with new fraudulent strategies.
One of the world’s top insurance companies AXA has collaborated with UK-based startup Darktrace to combat sophisticated cyberthreats. Given AXA’s central role in financial markets, cybercriminals’ motivation goes beyond personal financial gain and concerns global financial stability and politics.
During the pilot phase, Darktrace’s Enterprise Immune System embedded in AXA’s main network started learning the common patterns of every user. In the long term, this system managed to detect suspicious activities before they escalate and become serious threats. After the successful pilot, AXA deployed Darktrace AI system across its entire network, which now autonomously monitors and reacts to a range of advanced cyberthreats. Moreover, AXA’s cybersecurity team has access to comprehensive reports on these threats and can adequately react to the most pressing ones.
We’re not being attacked by human beings anymore. Computers are attacking us, software is attacking us, the only way forward is using artificial intelligence.
The aforementioned case is among the most advanced cybersecurity solutions on the market. It’s specifically trained and tailored to a particular business model, hence its price goes way above average. However, those interested in AI for small business can deploy pre-trained solutions like MalwareGuard developed by FireEye, which would be sufficient to withstand many common fraudulent attacks. FireEye uses an AI-based system to protect networks from various known web-based threats including email malware and remote system commands. Despite the simplicity of its deployment, to unveil the full potential of this solution, hiring data science consultants would most likely be required.
WANT TO START A PROJECT?
It’s simple!