Artificial intelligence adoption across the insurance industry is taking off. In its report, “The Dawn of the Age of AI in Insurance,” Genpact noted that insurers recognize the value and potential of AI, “with 87% of carriers investing more than $5 million in AI-related technologies each year. This is more than both banking (86%) and consumer goods and retail companies (63%).”
Genpact’s researchers also projected that AI would make a fundamental impact on the insurance industry in the coming years. Signaling this prediction will hold, the National Association of Insurance Commissioners’ Innovation and Technology Task Force formed a working group devoted to AI, which drafted the recently adopted “Principles on Artificial Intelligence (AI)” to help guide third parties, regulators and NAIC committees looking into AI in insurance.
As the technology continues to mature, carriers continue to experience better outcomes and greater operational efficiency. And the really exciting part is that we’re still in the early innings in terms of what AI can do as well as the types of that are possible.
So, where will the industry go from here? What do organizations need to do to prepare for next-level insights? Let’s dive in.
The Data Revolution
AI is only as good as its data. It’s a phrase you’ve probably heard before. From my perspective, this means organizations need both accurate data and a high volume of data. What is fed into models must be as clean as possible if you want to make predictions and yield insights with any degree of confidence. Yet, if the amount of data fed into the system is too small (a couple hundred claims vs. 1,000,000+), your results will not be representative either.
Data accuracy and volume become even more important as the nature of claims changes. Based on our studies, new claims include factors that adjusters and carriers haven’t seen before. More precisely, 34% involve a rarely visited provider while a quarter involve comorbidities that complicate claims. Additionally, nearly all claims involve claimants the organization is meeting for the first time, with their own medical history and personality.
These issues are prompting a shift in what data is used in AI systems. In addition to incorporating new types of data, such as medical record images, claim notes, and language processing, organizations are also beginning to pool claims data across the industry.
New mechanisms are being developed that de-identify and anonymize claims from participating carriers so that they can be used in aggregation in the AI process. Doing so enables organizations to safely share valuable information, as each carrier benefits from the cumulative treasure trove of data. Our cross-carrier analysis of claim predictions already show that accuracy rises to greater than 90% when data is pooled, compared to 70-80% for machine learning predictions within a single carrier and 40-60% for adjusters working without the use of an AI system.
There are other pluses to data pooling as well. Organizations are able to make initial predictions where none were possible before. Insights can shed light on where a carrier might want to extend its offerings to include a new line of business, or perhaps they will help create new industrywide benchmarks, such as reasonable settlement ranges for particular types of claims. Cross-industry data may also provide a clear view of providers’ and attorneys’ actual performance as viewed across the broader lens of industrywide claims. These are all enormous benefits that will help AI systems meet the greatest needs of both individual carriers and the larger insurance industry.
Change Is Coming
While the idea of collected insights is tremendously exciting, there can be some initial unease that accompanies data sharing — even when it is de-identified and segregated. To this I would say that it’s true other participants will have access to your data, but it’s also true you’ll have access to theirs. This creates a win-win situation similar to international trade. Countries may be competitors, but in the end, it is better for them and their people to cooperate.
When it comes to sharing data, what an organization gains is the ability to see far more phenomena than they would ever capture on their own. Even the biggest carriers have less than 10% market share, so by participating in a cross-industry data collective, they gain an enormous influx of relevant insights that improve their outcomes and make them more effective.
There is no sharing of proprietary information, only the potential to understand things such as how well providers address particular injuries or how certain lawyers perform. This is particularly important during a period such as the current pandemic, when changes in claim trends surface in very short order. Having the best, most current information allows organizations to spot such changes faster and adapt accordingly, which saves time and money over the long run.
But there is a narrow line to walk for this vision of shared data to be successful. A balance must be struck between accessibility and security. Data is an organization’s most valuable currency, and as such, security has to come first.
Security in Sharing
When looking for the right partner to build an AI system and effectively share the resulting information in a data collective, processes that ensure security should not be viewed as trivial. Any vendor should have implemented security best practices and controls to achieve compliance and certification with the most widely adopted security frameworks in both healthcare and IT. Some of the certifications to look for include the basics like Service Organizational Control 2 compliance to safeguard data, as well as Health Insurance Portability and Accountability Act compliance, which secures patient data along every step of the care process. Many companies are adding certification from the Health Information Trust Alliance to demonstrate they’ve achieved the highest level of HIPAA compliance in order to make customers feel more at ease.
It’s also increasingly important to show compliance with regulations like the General Data Protection Regulation (GDPR), which guarantees data protection and privacy in the European Union, as well as the California Consumer Privacy Act (CCPA) that offers California-based consumers the right to access their data, delete their data, and stop the sale of their data to a third party. These are essentially table stakes today.
Given the direction of the industry and the significant efficiencies and advantages of AI in claims operations, it is only a matter of time before all carriers begin sharing their data to get the most out of their AI investments. The change will not happen overnight, but it is inevitable. With the AI revolution underway, no one wants to be left behind. As such, carriers would be wise to begin plotting their data accessibility and security strategies.
As first published in Claims Journal