The journey of a commercial claim with AI
Emerging tech is helping to make massive loads of claims data more useful for adjusters to enable better outcomes.
Across commercial insurance, from auto liability to workers’ compensation, claims that involve a bodily injury tend to follow a common lifecycle from the moment a claim is opened until the date it is resolved. As claims evolve through stages in this journey, a growing set of data gets added to the claim file, pulled from bills, notes, images, payments and other files. For complex claims, the journey can extend for years and involve updates and files from adjusters, nurse case managers, providers, attorneys and others. When AI is added, the data in this journey becomes visible, forming a more predictable trajectory that enables adjusters and supervisors to effectively manage the claim, reduce costs, and enable better outcomes for all involved.
Traditionally, much of this data is not fully used, either stored beyond the reach of claims teams or overlooked by busy adjusters just trying to stay on top of their caseload. But now, image and language processing can aggregate and organize that data and machine learning can be applied to generate insights and predictions about the likelihood of future events. All of that gives adjusters a powerful new tool to help claimants get back to health faster — and reduce costs for all parties involved.
To see the potential, let’s look at how a claim in a specific line of commercial insurance — workers’ comp — evolves today. A person gets hurt on the job, and their employer files a claim. A claims adjuster collects notes from the worker and employer and keeps track of progress as the claimant visits a doctor. The worker gets additional treatment over time, which is also documented, and then goes back to work if all goes well; if it doesn’t, they may enlist a lawyer, which extends the life cycle many months or more. Much of the data collected is never analyzed and often only a fraction is reviewed. Even if they had the time, adjusters would struggle to draw disparate connections between all of it.
This is how things have worked for decades, but they haven’t actually worked, at least not to their potential. Claim costs have risen due to using reactive rather than proactive management and tools, allowing 28% of claims to drive 80% of claim costs.
To make the mountains of information that accompany a claim more useful, artificial intelligence (AI) is slowly being introduced into the claim’s journey, starting with predictions that are most critical, such as claim costs or attorney involvement. But AI is still very new to insurance, and today’s claims teams are only scratching the surface on how it can be applied for the betterment of all constituents.
AI in action
Let’s take a look at how AI can reshape the journey of a claim. In this new vision, the pivotal points in the claim life cycle remain the same, but with data science mixed in, the journey becomes smarter and more efficient. A simplified flow may run as follows:
Intake and Phase I predictions
This is the area where several vital applications of data science emerge. Phase I also sets the stage for each subsequent phase.
A claim is opened and all of the initial details on a worker’s injury are placed in a cloud-based system. As such, relevant case details are accessible to analysis; no one needs to go digging through case files manually or waiting for hard copy images to arrive. With advances in AI, asynchronous data sources, such as notes and images, are incorporated so that the system can understand and use them to establish trends and make connections.
Using data science, predictive algorithms assess not only the data collected on the claimant, but they also can compare it to all of the data available in the system on other claims — which translates to a repository of millions of data points. This enables systems to do things like assign claimants to available providers who are believed to be the best equipped to treat them and flag cases, which might warrant extra care.
With relevant information available at claims adjusters’ fingertips, AI systems eliminate significant amounts of time spent by adjusters working on each claim. For example, the potential for the claims rep to simply direct a claimant to a provider she knows — whether or not it is the right provider for the case — dramatically decreases if there is an algorithm that looks for available specialists within a 30-mile radius of the claimant, who have a track record of positive outcomes, and who are in-network and available. That information allows the rep to instantly choose the best provider for the job. This is particularly important today, given the limitations on care access imposed by COVID-19 and the new types of cases that are appearing in claims. Even veteran adjusters can’t rely on their same roster of thoroughly vetted providers; they need a system that can identify new possibilities that will deliver desired outcomes.
Well-trained machine learning can identify potential markers for problems within a claim based on data from thousands of other claims. With intelligence, systems can flag these cases for adjusters with recommended actions (again, based on data) so that adjusters can intervene early. Remember that 28% of claims result in 80% of claim costs stat? Early intervention can diminish this issue and save companies millions of dollars a year by being proactive with these claims; AI-based systems make this remarkably fast and easy to do.
In each of these cases, the adjuster is saved from countless hours of research that can never be as complete as what a machine can turn out in only seconds. As a result, the adjuster has more time to devote to the cases that matter most. She can also handle a greater number of cases in order to help more people more quickly.
Care delivery and Phase II predictions
In the next phase, the claimant goes on to receive additional care, and data related to the claim is continuously monitored. Algorithms can flag coding errors that prove costly throughout the life of a claim or look for signs of provider fraud. A wide range of claimant sentiment signals can also be triggered by the continuous data feed, including potential return to work concerns, comments that often lead to litigation, and treatment concerns with the primary doctor. The signals are time-stamped, and together, they form milestones in the claim’s journey. The sequence of adjuster contacts and recommended actions can also be plotted.
If all goes well and no flags emerge during the care delivery phase, the claimant heals and returns to work. The claim is resolved efficiently. In some cases, though, the claimant may pursue litigation, which can add thousands of dollars and several months onto the life of each claim.
Litigation and Phase III predictions
Claimants might involve attorneys soon after filing a claim or much later in the process. In either situation, attorney involvement represents a whole new phase in the claim’s journey.
With the right algorithm and the means to pull data, AI systems can identify the best attorneys for a case based on outcomes. They can also pull and analyze key settlement data so that organizations can make the right decision about whether to settle and when. As a result, cases are resolved as quickly as possible, and claims are closed more efficiently than ever before.
Conclusion
While this is only a snapshot of how AI can be applied today, it illustrates how having actionable, contextual information in your hands at the time you need it, particularly early in the claim’s life cycle, enables claims to be handled efficiently, resulting in better care for claimants, lower costs for organizations, and fewer frustrations and manual tasks for adjusters.
As the world of claims continues to evolve, this technology will become even more important, driving new insights and breakthroughs. Increasingly, AI produces results that empower organizations to adapt faster and more responsively. It’s exciting to think about where it will go next.
Ji Li, Ph.D., is data science director at CLARA analytics, has leadership responsibility for organizing and directing the CLARA data science team in building optimized machine learning solutions, creating artificial intelligence applications, and driving innovation. Dr. Li is well-published in fields related to computational theory and big data applications. His specific expertise and interests include machine learning, deep learning, text mining, and natural language processing and understanding. Dr. Li received his Ph.D. in mathematics from the University of Connecticut.
Related: