This article is part of a VB special issue. Read the full series here: The quest for Nirvana: Applying AI at scale.
For Aflac, which provides supplemental insurance to more than 50 million people worldwide (and is well-known for its duck mascot), delivering AI at scale across the organization has become a top priority since the pandemic.
Aflac has been forced to accelerate its digital transformation, including artificial intelligence (AI), as the pandemic severely challenged the company’s traditional in-person, independent agent/franchise business. The trick, however, has been choosing the best AI use cases among competing priorities, says Shelia Anderson, who joined Aflac as CIO last July.
“We’re thinking about the business challenge and outcomes we’re looking for,” she told VentureBeat.
When it comes to AI and machine learning, that includes focusing on the overall viability and desirability of the company’s models, and asking questions such as: Is the model needed by the business? Is it solving a specific business need? Does the company have the technical solutions it needs? How long will it take for the model to bring value to the business?
A clear opportunity to automate claims
Aflac has long had a focus on big data. Today, Aflac supports agents and brokers with AI and ML models that aid in suggestive selling, flagging at-risk accounts and identifying dormant accounts that are candidates for reactivation.
But developing a solution that could scale AI across the organization has been a high priority since 2020, said Anderson. Just last year, the company rolled out what Anderson calls its first significant AI-driven platform that uses AI and ML to transform how Aflac processes claims.
The platform consists of a set of models trained on business rules tailored to the company’s various product lines. The goal is to automate routine processes, allowing the company to pay claims more quickly.
There are three main components to the platform:
- An AI-based document digitization pipeline to automatically extract, classify, annotate and index proof-of-loss documents
- Knowledge graphs to map extracted information from documents for a better context of processed data.
- An end-to-end, AI-based claims processing workflow for adjudication across different lines of business, allowing for fully automated or assisted, error-free, human-in-the-loop processing.
“This helps our customers to be adjudicated faster and with more accuracy,” Anderson said, pointing out that before the AI solution was implemented, about 46% of Aflac claims were not fully automated.
Aflac has many different claim types, she explained, but one of the first clear opportunities to scale AI was around the company’s wellness benefits. These are included in most of its accident, hospital indemnity and cancer insurance policies. Essentially, Aflac pays customers money for getting yearly checkups and medical screenings such as physicals, dental exams and eye tests.
It turned out there was a high volume of lower-dollar payout claims requiring time-consuming customer interactions.
“For simple claims that don’t require proof of loss, like wellness claims, we want to pay out quickly,” said Anderson. This “allows our customer care specialists to take care of our policyholders [who have] more complex situations.”
Scaling the AI platform
Now, Aflac is working to scale its claims automation platform to other types of claims.
“The benefits that the business case has proven are improved customer ease, reducing our pain points through the journey, and increasing our touchless claims, which was a benefit to our internal workforce as well as our claimants,” Anderson said. “Streamlining with a rules-based AI reduces error rates and frees up our resources so they can focus on more critical claims where people may actually need to hear a voice on the other end of the phone, maybe dealing with more severe health-related issues where that personal touch is needed.”
Anderson said she believes Aflac has only just hit the “tip of the iceberg” when it comes to implementing the platform. She has plans to expand the same capability across the organization in 2023. That, she pointed out, is the value of getting a model that works well, one that solves a basic challenge and takes advantage of an opportunity in the marketplace.
“You can take that and stamp it across your other lines of business with a similar problem,” she said. “So we’re taking this and expanding it in our accident and hospital lines of business, and we’re also adding other capabilities in the future around cancer, dental and vision.”
In addition, she added, there is an opportunity to extend these AI capabilities beyond the claims process, to any use case that needs to be automated based on prediction.
Aflac’s biggest AI scaling challenges
Besides prioritization, one of the biggest challenges in scaling any AI effort across Aflac is getting participation from various organizational entities, Anderson said.
“For example, our partner that runs the analytics side of our business has a front-end team,” she explained. “We have a back-end data team and then we have business teams that we work with as well. So managing and prioritizing across that ecosystem, whether it’s AI or whether it’s another business initiative, that’s always going to be something that is a challenge for us.”
In addition, in a high-demand space like AI and machine learning, attracting and retaining talent with the right skill set is a major challenge. “It’s something we all have to stay laser-focused on,” she said.
Applying AI to improving customer retention
Overall, Aflac’s claims automation platform has helped with customer service and customer retention, Anderson said.
It’s about “how we spend the time that we need for those highest-priority customers and claims while automating others,” she said. “I think that customer service is going to be key in leveraging AI in the future.”
That said, she added that she believes allowing some AI capabilities to mature has been an important part of Aflac’s journey — taking time to make sure it doesn’t take needless risks with customer interactions.
“If you want to be first to market with something, of course, that’s just a risk you’re going to have to take,” she said. “But for Aflac, I believe that allowing some of these capabilities to mature was definitely part of the journey.”
VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.