The unique potential of ML in transforming the insurance value chain
Historically, the insurance industry has majorly been described as conservative in its responses to digital disruption. However, what started off in the pre-Covid-19 times as a campaign to convert some incumbents into digital-first players has come across as a shot in the arm in the post-Covid-19 scenario.
Insurers who were holding back due to voluntary and involuntary reasons are realizing they don’t have the option of status quo anymore. In short, insurance is waking up to market realities. Companies are realizing that customers are demanding personalized products, better service, and unique possibilities from them, especially in the post-Covid-19 world. If they can’t deliver those, others will. And that surely is no way to be a part of the competition.
What comes across as a veritable promise for insurance (3.0 if you will, in the post-Covid world) is — no prizes for guessing — technology. It’s no secret that making the best use of new-age technologies will enable insurers to constantly innovate, ramp up customer satisfaction, and stay ahead in the race. In other words, going digital is not a good-to-have for insurers anymore; it’s the new normal. We’re talking about digital touchpoints, digital offerings, digital claim settlements, digital damage analysis, digital customer support, and the likes.
Among technologies, leading insurance players and start-ups alike have discovered the enormous potential of machine learning (ML) in transforming their business, and the sector as a whole. Machine learning — a subset of AI that’s all about learning from, and making predictions based on, data and past experiences — itself is poised for an exponential growth with Fortune Business Insights predicting the global ML market to reach $117.19 billion by 2027, growing at a CAGR of 39.2%. This is a favorable time for the coming together of ML and insurance. Here’s why.
Factors that make ML a perfect fit for insurance
Data explosion: Data has always been at the heart of the insurance industry. Moreover, the rise of smart and connected devices is resulting in a new avalanche of data. But most insurance companies are able to process only 10–15% of accessible structured data. Now, ML is bringing in new opportunities to train and derive actionable insights from the humongous amounts of data out there — whether structured, semi-structured or unstructured.
Open source ecosystems: With the evolution of technology and omnipresence of data, systems are also evolving in an open source manner to ensure data is shared and used securely across all industries. For example, insurance can gain access to your health or driving data and leverage ML to customize your health insurance policy. For example, State Farm’s Drive Safe & Save uses 2D dashboard camera images of drivers to derive insights and customize insurance offerings and offer customers discounts based on their driving behavior.
Rise of cognitive technologies: The growth of ML is propelling deep advances in natural language processing and image recognition techniques. And insurance, in its quest for greater customer satisfaction, is utilizing them to offer novel services ranging from chatbot support for agents to on-the-spot damage analysis by drones for customers. The shift is from a ‘detect and repair’ model to a ‘predict and prevent’ one.
Critical insurance areas impacted by ML and key use cases
Driven by ML, the insurance value chain is turning richer with innovative capabilities.
Claims management: Insurers are using ML to build non-static and complex decision trees that can help them automate underwriting, pricing and claims processes. That results in enhanced customer experience through efficient management, focused approach and investigations, better understanding of claims costs, and, of course, reduced claims settlement time. Tokio Marine’s OCR- and ML-led claims document recognition system handles handwritten claims documents with a 90% recognition rate, resulting in 50% reduction in input time and 80% reduction in human error. Tractable’s agents can upload images of a damaged vehicle on the platform along with an estimate based on the damage. The platform analyses the inputs vis-a-vis thousands of stored images and related payouts, and alerts the agent if the current payout exceeds the norm. This way, it saves a lot of loss related to claims leakage.
Fraud detection: In the US, fraudulent claims account for $40 million annually, and in the UK, 350 fraud cases are uncovered everyday. ML helps insurers analyse large quantities of data and identify fraud claims faster and more accurately, flagging them for further investigations by humans. Shift Technology uses ML to find fraud patterns in deep claims data sets, which can then be applied to claims to identify potential instances of fraud. Similarly, Lemonade’s AI chatbot, Jim, matches a user-submitted description with similar descriptions stored in its database to detect and prevent fraud.
Underwriting: Insurers are leveraging ML to not only study personal data (like driving or health data) to design personalized policies. They are also using the technology to predict premiums and losses, and manage risks better to help underwriters stay on top of their game. For example, UK-based O2 Drive car insurance leverages an in-car device to track driving habits and regulate premiums based on them. The safer the drive, the better price one gets. Likewise, Daisy Intelligence’s software helps insurers automate their underwriting experience through ML. They provide cost suggestions to individuals based on their risk factors like age, location, comorbidity, etc.
Customer support: Thanks to ML, a quick review of customer profiles enables auto-generation of personalized policies and advice. According to a survey, 74% consumers feel happy to receive computer-generated insurance advice. Chatbots that work with messaging apps are also giving out answers to simple insurance queries and resolving claim queries. All State’s ABIE chatbot, powered by NLP and ML, handholds agents, business owners and consumers with top-of-the-mind questions and step-by-step quoting and policy issuing guidance. Processing 25,000 inquiries a month, ABIE is able to automate a hitherto human-driven service that led to “long wait times” and “lost business opportunities”.
Damage analysis: Human agents assessing property damage to fix disbursal amounts, settle claims or reject them are passe. Trust computer vision-powered drones to do a quick aerial survey of a crop failure. Or even a smartphone camera to assess damage to your car that met with an accident. These tools are trained with thousands of images of similar past incidents and their reimbursement trends. Liberty Mutual’s app lets users assess the damage to their car in real-time in case of an accident. The ML-powered app is trained with thousands of car crash images that enables it to quote a repair cost estimate on the spot. Ping An’s “Smart Fast Claim” uses image recognition and pricing algorithms to recognize auto damages, improving claims efficiency by more than 40%.
Change your bottom-line powered by ML
Machine Learning as a subset of AI always had a huge potential. Global situations like the pandemic are a blessing in disguise for insurers to leverage the technology and make the most of it that impacts your bottom-line. No wonder that apart from the key areas mentioned above, ML is also making headway in other facets of insurance like quality lead prediction, marketing, customer retention, and so on. As machine learning keeps “learning”, its scope will broaden, allowing highly skilled and experienced industry partners to enable insurers in taking their offerings and services to the next level.
A version of the article appeared here.