Sergey Nivens - Fotolia

Machine vision makes paper a thing of the past for insurers

The insurance industry is buried in paper-based processes. Former MetLife CIO Gary Hoberman aims to change that with a platform that runs on AI and machine vision.

Can a notoriously slow, paper-based industry rejigger itself for the digital age? When it comes to insurance, an industry rich in data but a laggard in digitization, a software-as-a-service startup founded by former MetLife Global CIO Gary Hoberman believes a solution is at hand.

Unqork Inc., which bills itself as an insurance fulfillment platform, is using a host of artificial intelligence technologies to optimize, digitize and automate arcane and inefficient processes.

Formed last year, the 15-employee startup isn't betting on flashy customer-facing AI applications. There are no chatbots or virtual assistants; no image recognition apps that can accurately distinguish a dog from a blueberry muffin.

Instead, the Unqork platform lives under the covers of the sales processes, according to Bassam Chaptini, chief analytics officer at Unqork and former McKinsey partner in the data and analytics practice. One of the most tangible examples of how Unqork is using AI to make insurance processes more efficient is by applying machine vision -- a set of technologies used to extract information from an image -- to turn paper-based applications into digital assets.

Bassam ChaptiniBassam Chaptini

"People have tried digitization a lot -- it's not for lack of trying," Chaptini said. "There's just been no appetite for adoption because, for decades, agents have sent paper applications to the carriers."

But tradition is wearing thin. Between sending back applications that haven't been filled out properly and verifying data, the back-and-forth between the agent, carrier and customer can seem endless, Chaptini said. Specifically in the life insurance business, almost 65% of all applications are still paper-based. On average, he said it takes nine exchanges between an agent and a customer in as many as 40 days to complete an application.

Digital failure and missed revenue

Ellen CarneyEllen Carney

The life insurance industry is ripe not only for disruption, but for fixing, a phrase Unqork CEO Hoberman used to describe the company's mission to attendees at the Pacific Insurance Conference last fall. Analyst Ellen Carney, who covers e-business at Forrester Research, agreed, asserting that life insurers are seven to eight years behind in digital capabilities compared to auto insurers.

Carney is the author of a soon-to-be published benchmark study of the top life insurers in the United States and Canada on their digital sales functionality -- from discovery to research to the application process.

While life insurance companies successfully attracted customers to their websites and provided basic information, Carney's research showed they fell down at explaining how much coverage a customer needed, the process of underwriting and connecting customers to the right advisor. In one case, an insurance company suggested Carney meet with an agent who was more than a four-hour drive away. "I'm not kidding," she said.

In other cases, insurance companies didn't provide a convenient phone number to call if a customer had a question; didn't follow up on abandoned shopping carts, even when credit card information was entered; or disappeared altogether. "We had situations where we asked to have an advisor reach out to us," she said. "You never got to pick who the advisor was. And no one ever called you."

The digital floundering by life insurers leads to missed revenue opportunities in a growing market. Carney said interest in life insurance is up and that one-third of customers apply for life insurance spontaneously. "Yet the experience has not caught up with how we book an airline flight, how we open a checking account and how we order something from Amazon Prime," she said.

The way forward won't be easy for a segment of the insurance industry that is still making up for ground lost during the global recession and simultaneously under threat from "insurtech" startups aiming to disrupt its business model, Carney said. But the industry clearly recognizes that it needs to invest in technology. According to her benchmarking studies, life insurers are investing 14% of their IT budgets on emerging technology. For comparison, property casualty insurers are investing between 7% and 8%.

"They're playing catchup with new kinds of technology," she said. That includes -- or should include -- adopting artificial intelligence tools like machine vision. In a January report "Accelerate Your Digital Insurance Strategy with Six Emerging Technology Building Blocks," Carney recommended that insurers invest heavily in artificial intelligence, including a set of technologies she labeled AI and cognitive as well as deep learning.

Ninety-five percent automation

That's where Unqork is hoping to come in.

Unqork's automation tools are designed to "shrink that time to issuance to the benefit of the customer, agent and the carrier," Chaptini said. Unqork takes a scan of the paper application, which includes 500 data fields, uses machine vision to extract the data and then populate its electronic application or e-app.

The machine vision algorithm, which, Chaptini said, advances in deep neural nets have made more precise than ever before, can extract about 95% of the data accurately. A separate team, which Chaptini referred to as a "concierge service," double checks instances where the algorithm indicates a lack of confidence.

"That's done by humans," he said. "We will push automation as much as we can, but we also want 100% accuracy."

Unqork: What's in a name?

At the Pacific Insurance Conference, Unqork CEO Gary Hoberman said he originally wanted to call the company Uncork, as in unbottling a company's competitive advantage. But to buy would have cost $500,000. "The 'q' was $14.99 on GoDaddy," he said, "and we are all about value."

The startup also validates the data, using APIs to check the application against information sources such as Experian or Facebook to double-check that names, social security numbers and birthdays are accurate, and it flags any data fields that are missing or incomplete. It then sends an insurance agent a text "in under a few minutes" with a link to an interface that details what's needed to complete the application, according to Chaptini.

"What we see is a shrinking by 40% in terms of cost to the carrier to process the application, and a dramatic shrinking of the time-to-issuance, which means faster commissions for agents," Chaptini said. That has a ripple effect on improving process by providing agents with an incentive to rely on the digital application and by cutting down on customer churn. "Data has shown that the longer you wait for an application to be issued, the higher the rate for churn for that particular customer," he added.

Unqork is also utilizing AI to automatically match a customer with the most appropriate agent, which Chaptini said can lift sales; and, while it's not being applied to the underwriting process directly, it is being used to analyze and optimize how the underwriting is executed, a way of looking at the process without bumping into regulatory constraints.

"Life [insurance] gets particularly complicated with a product that could have as many as 25,000 rules in terms of the underwriting process," Chaptini said. Unqork digitizes the rules, some of which are in an underwriter's head, and then provides insight into what rules are or aren't used and how they correlate to the underwriting outcomes.

"It's an opportunity for a carrier to dramatically simplify their products, their rules, and to make empirical decisions," he said.

Dig Deeper on Machine learning platforms

Business Analytics
Data Management