Big Data World Asia: Machine learning in insurance, with Munich Re’s regional head of data analytics
While machine learning has transformed many industries over the past decade, one area that is still playing catch-up is insurance. It’s a sector used to finding itself trailing behind other industries’ tech adoption, where high running costs of legacy systems squeeze budgets to such an extent that it’s hard for firms to stump up the cash necessary for driving innovation. While online comparison services have proliferated in recent years, signing up to and managing the policy invariably involves the pens, paper, and printers that other digitally-transformed industries have long since left behind.
But as data systems have matured, and AI, deep learning and connected devices continue to proliferate, the insurance industry is growing up and starting to harness the numerous benefits AI and ML can offer. Like all other industries with growing stockpiles of data, models can be deployed to dive deeper – transforming toothless data into tools of automation and prediction.
Machine learning models have proven adept at speeding up functions, for instance expediting claims that meet defined criteria. Natural language processing is also being used to derive insights from text data in claims forms, social media, emails or chats. And data from connected devices is being used to improve decisions about health policy premiums — or in the case of telematics data, deliver better car insurance rates and prevent fraudulent claims.
Just as in other industries, disruptive first-adopters are emerging as market leaders. Despite the rising number of them entering the space, such as ML-based insurance policy provider Lemonade, traditional investors are also getting in on the act. According to Genpact, 87 percent of insurers are investing more than $5 million in AI each year.
Many in the sector are optimistic that they can turn their tardiness into an advantage by learning from the failures and successes of other industries. It’s important not to lump all insurers into the same basket. In fact, there is a spectrum of adoption. Some companies are still deciding whether to invest in a data warehouse, while others, like Ping-An, are already using AI to settle millions of claims a year.
One such insurer that probably sits in between these two poles is German-based Munich Re, which has steadily increased its investment into big data and analytics in recent years. Ahead of his appearance at Big Data World Singapore on 9 October Techerati spoke to Dr Vishnu Nanduri, who heads up the Data Analytics program for the company’s Non-Life reinsurance business (for SEA, Japan, Korea, and India), a type of insurance that reduces risk by passing on a portion of a policy’s liabilities to another provider.
I ask Nanduri to explain the explosion of innovation in insurance. He credits digitalisation and the rise of ecosystems with enabling automation and cross-platform data sharing that, in turn, have made the components of the insurance value chain faster and more efficient.
“The days of long and tedious underwriting processes, painful claims hoops that a customer has to jump through to get paid, paper forms to buy insurance; will very soon be things of the past,” he says. “The future of insurance will be a maximisation of digital, minimisation of human touch, and increased use of AI and ML in many functional areas.”
Nanduri is most excited about AI’s impact on three areas of insurance: Underwriting, claims processing and customer engagement.
In insurance, underwriting refers to the act of receiving money for the willingness to pay a risk, whether that’s the risk of a car crash or the need for surgery. Underwriting is essentially a mathematical calculation of the likelihood of an event occurring, so it’s not surprising that it was one of the first areas where insurers looked to deploy AI models. The major change that is occurring, however, is that rules-based automated underwriting platforms are being enhanced by machine learning to make the process more streamlined, efficient, and accurate, Nanduri says.
Surely the area of insurance that has most to gain from AI is claims processing — what Nanduri calls “the rubber-meets-the-road’ moment for an insurer. A claims experience typically determines whether customers are retained or lost. Getting claims right improves the rate of renewal, customer loyalty, and inevitably catalyses additional purchases. Getting it wrong leads to the opposite and thus the deterioration of a customer’s lifetime value. Aside from being the central cog upon which an insurance company spins, Nanduri points out that claims processing is also an area where manual processes remain rife. With such room for improvement, the benefits of incorporating AI could likely exceed those felt in other areas and industries.
“Efforts are now afoot in many insurance companies to digitalise the claims process and then automate portions of claims decision making based on model recommendations,” Nanduri explains. “Ping An’s image-based claims damage assessment is often considered state-of-the-art in the area of claims management, by many.”
Finally, using AI to target marketing to the right individuals will make insurers more successful at attracting new customers. Targeted marketing is not unique to the insurance industry but the pace of adoption is slower than some other industries. Nanduri says it’s still typical for some departments to rely on excel based models to cross and upsell, but that sophistication has increased markedly over the last few years. Some insurers are deploying deep learning and other advanced machine learning methods to identify purchase patterns across hundreds of variables, he says.
“Offering the right customer, the right product, at the right time, via the right channel is how insurance companies are trying to engage customers positively,” he says. “Calling digital natives on their cell phones versus a customised product recommendation on their social media homepage can make a world of difference for an uptick in sales.”
Despite the great strides made in these areas, Nanduri identifies several challenges that need to be overcome before adoption increases and delivers on its potential. Among them are leadership mindset, data quality, data availability, data storage and use, talent management, education and training, and culture. Don’t miss his presentation at Big Data World Asia to hear his insights on how these bridges can be crossed.
“Companies that are successful in AI and digital transformation get more of these components right than the others that are not as successful,” he says.