Canadian Underwriter
News

Look for AI to “hyper-personalize” customer experience in 3-5 years: Aviva exec


July 20, 2021   by David Gambrill


Print this page Share

Hyper-personalization of customer experience in three to five years’ time is a big reason why the property and casualty insurance industry is ramping up the use of artificial intelligence and machine learning now, an Aviva Canada tech executive suggests.

“I think customization and simplification of customer journeys is going to be very critical,” Baiju Devani, senior vice president of data science and chief data officer at Aviva Canada, told Canadian Underwriter in an interview. “We are very early on in that journey.

“We talk about hyper-personalization [of the customer experience], but very little of that actually happens [right now]. From a technology point of view, it’s a big task. And that’s what I think the future three to five years from now holds for us.”

What is meant by “hyper-personalization?”

Where data is concerned, the P&C industry has traditionally operated on the basis of statistical averages. For example, in auto insurance, insurance companies typically collect large amounts of consumer information, including the age, sex, and home and work addresses of the driver. From this data, they can estimate the average likelihood that an insured driver will crash the car. Or in home insurance, they can figure out how often a property in a certain area typically floods, and estimate the average cost of a flood claim.

By personalizing the insurance product and experience, the industry is seeking to get to a place where the customer’s insurance will more accurately reflect their precise individual risk characteristics, as opposed to guessing based on historical averages.

AI and machine learning are the way to get there, said Devani. “To a large extent, we are in the game of averaging, but the less we do, the better it is for our customers.”

He cited specific examples of how AI can lead to more personalized experiences in the areas of customer service, insurance product pricing, and claims service.

Regarding customer service, Devani noted how AI can help market information to a consumer’s specific risk profile.

“We have a wide range of products, and we’re not tailoring [the information about those products] to the customer at that moment,” he said. “There’s a lot of a lot of hyper-personalization we can do, saying that, ‘Hey, this looks like this is a younger client. Maybe we want to load our pages with a lot more educational content so that the coverages, the endorsements, and all those things are much more prominent.”

In contrast, maybe the client is older and has a more sophisticated knowledge of insurance. In that instance, they might need something less information-based, and more price-focused. “The focus would be on packages [available], as opposed to education,” Devani said.

Ultimately, the name of the game for insurers is “personalizing the content I’m giving you at the point of sale, because I know you [as a customer] might be at different experience levels,” as Devani explained. “That’s an example of personalization. You can keep pushing the limits on that. That’s not something to do with insurance, per se. That’s just the buying phase.”

But AI can help to improve the underwriting phase as well. Machine learning can help automate repetitive tasks, for example, leading underwriters to look at more complex data patterns and outputs. For example, automating geographic segmentation will free up time for data analysts, actuaries, and underwriters to focus on other, more complex data patterns.

Devani suggested the use of AI to help underwriters price risk remains a work in progress.

“We all see the value in telematics,” he said. “It’s the stuff that’s simple, like miles driven. That’s easy. We know how to use that. But when we think about AI for pricing, that’s a little bit harder to define mathematically and apply [in a way] that’s going to improve a customer’s experience or improve [a company’s] bottom line. That’s another interesting dimension.”

In claims, examples of AI often focus on fraud detection. No doubt, AI can help identify attempts to defraud insurers. A simple example is detecting the same address for claimants, or multiple billings for by service providers, appearing in multiple claims events over a short period.

But AI can also help the claimant recover more quickly from a claims event at a lower cost, Devani said.

“Our adjusters are interacting with an AI that is recommending an outcome based on the severity of the accident, where you [as a driver] have been, what you’re driving, and who you are,” Devani said. “AI might suggest to an adjuster: ‘Perhaps this isn’t a total loss, you should be recommending a salvage to the nearest shop.’ Or, if it’s a repair, where you send that repair.

“Those are the little things that we do that add up to a better customer experience because they are more focused on you [as a customer] and get back to your [pre-accident life] as soon as possible.”

 

Feature photo courtesy of iStock.com/Jirsak