Claims processes can be easy to overlook – until a crisis arises. Perhaps that’s why COVID-19 has suddenly refocused attention on claims management.
Research shows that timely, efficient, accurate, and fair claims settlements can strengthen the customer relationship in profound ways. On the other hand, underinvestment in claims can result in real reputational damage, higher operational costs, and rising vulnerability to claims leakage, including fraud, waste, and abuse. A survey by RGA, conducted just prior to the pandemic, showed that just one in four respondents (28%) used an expert claims system to assist in claims management and settlement, and fewer than 18% of those covered the complete end-to-end claims process.
While insurers in recent years have rushed to embrace underwriting digitization and acceleration to support top-line growth, it has been all too easy to overlook operational processes that can affect the bottom line. That is, until now.
Multiple technologies are emerging to address the need for greater efficiency in the less visible, but no-less-important, claims function. These fall generally within two areas of emphasis:
Virtually every claims process, regardless of product type, could benefit from digitization, but the details can make all the difference.
Take Robotic Process Automation (RPA). Traditionally, insurers have relied on large claims teams to manually retrieve, compile, and enter data from claims files. This time-consuming approach is increasingly incompatible with modern, just-in-time business practices and claimant expectations. RPA technology fast-tracks adjudication, empowering claims organizations to perform high-volume tasks at exponentially greater speeds.
How does it work? Do not expect androids to replace claims analysts anytime soon. Instead, RPA technologies simply manipulate existing software, such as claims applications or customer relationship management tools, using sets of predetermined rules. Optical character recognition and natural language processing systems can recognize and extract data from a range of documents, images, and other forms of claims evidence. This information can be compiled into centralized databases with higher accuracy, allowing human claims analysts to avoid repetitive policy servicing tasks and devote more attention to meaningful claims management activities. For example, in the past upon receiving a life insurance claim, the analyst confirmed the policyholder’s death through document requests; today RPA systems can scan public death records in milliseconds, updating claims files automatically. This not only relieves the bereaved from the burden of having to submit a physical death certificate, but also the claims analyst from having to retrieve one. Similarly, RPA systems can extract digital health data needed for a medical claim directly at the “point of sale” – from electronic consultation records held by a hospital or physician’s office. This can dramatically improve payment efficiencies, while removing the potential for error or fraud.
RPA is revolutionary for an even more compelling reason: incumbent claims systems present a serious barrier to automation. Most insurers depend on a web of interrelated technologies to handle policy administration, underwriting, payments processing, and more. Information is passed through complex networks, so introducing a new technology can prove dauntingly difficult. RPA systems eliminate this problem by working with and between legacy systems, so insurers do not have to install expensive infrastructure.
Such digitization has presented a powerful competitive advantage long before COVID-19 forced claims operations to go virtual. As a case in point, insurers in Australia years ago found themselves ill-equipped to manage a wave of total permanent disability products linked to generous benefits. One could draw a straight line between overwhelmed claims staff and rising consumer complaints. Automation, on the other hand, empowers claims analysts to anticipate and prevent congestion in the assessment process. Experience2 shows that employing RPA can reduce claims-processing time, eliminate points of friction between claimants and insurers, and even help companies cut adjustment expense, while maintaining and even improving accuracy. The resulting improved customer experience distinguishes insurers in a crowded marketplace.
Digitization has presented a powerful competitive advantage long before COVID-19 forced claims operations to go virtual.
Still, it is easy to overstate the power of automation. Claims management is an art, as much as it is a science, and as product lines grow more complex, the advantages of automation decline. For example, complex claims for critical illness and disability products are likely to continue to be handled by human experts. This is especially true when claimant interactions are required, often demanding nuanced judgment and consideration of psychosocial and medical factors.
Efficiency, in other words, is not the only consideration. Insurers cannot simply pay claims as swiftly as possible without risking unsustainable costs; these claims must be validated. That’s why claims analysts have the most high-stakes jobs in insurance. They must make accurate decisions while simultaneously limiting the time involved in doing so.
Enter augmented intelligence, a spectrum of approaches ranging from expert rule engines to exception reporting technologies designed to enhance, but not necessarily replace, human judgment. These systems help insurers triage more straightforward claims decisions while identifying and routing more complex claims in ways that guide the analyst. Ideally, everyone wins: the customer gets faster decisions, and the claim analyst’s extensive experience is put to more efficient use. Yet often these technologies are confused, overstated or conflated.
Consider the humble expert rules engine, the backbone of automated underwriting systems such as RGA’s AURA NEXT. The rules engine has been around for some time and relies on simple logic to match a code or condition with any exclusions identified by an insurer’s underwriting policy or guideline. And as common as these systems may be, they are not universal; many claims functions are still mastering or implementing engines to aid in decisioning.
In contrast, machine learning is often misunderstood. Typically, a data scientist “trains” a model to establish predictive patterns by feeding vast amounts of structured (organized) data into an algorithm. For example, past hospital lengths of stay for a given diagnosis could be used to project future outcomes. The model can help guide the claims analyst in making sophisticated decisions about leakage, delinquency, accuracy and fraud by applying a score to a relevant claim, but a human is most commonly required to teach the model by adjusting data inputs and the information typically is structured or homogenous to make sense to the system. It is also worth noting that these algorithms can be coded into expert rules engines to help improve performance.
Similarly, exception reporting involves the use of models and structured data to establish norms expressly to better spot outliers, such as an excessive length of stay or cost of treatment. A recent survey by Accenture3 suggests that insurance leaders increasingly believe such systems may hold the key to combatting claims leakage. Conventional wisdom holds that 3-5% of claims payments are inaccurate. Flagging disparities and delivering intelligence empowers human analysts to investigate anomalies.
RGA’s Automated System for Claims Leakage (RASCL) offers a powerful example of these augmented intelligence technologies in action. RASCL relies on a combination of published clinical rules developed by medical experts and predicative models developed by RGA data scientists to help identify fraud, waste, and abuse. It tests existing claims data against seven points of validation to identify problems in claims reporting or payment in real time. And RASCL is designed to integrate with, rather than replace, insurers’ current systems, avoiding any need for dual data entry.
RGA’s Automated System for Claims Leakage (RASCL) offers a powerful example of these augmented intelligence technologies in action.
Artificial intelligence (AI) is often used as a broad, general term but it has a very specific meaning. AI algorithms can include – but are not limited to machine learning– and can learn with less direction and far larger pools of data. These systems can independently sort through millions of claims files, examining all forms of information on millions of claimants to unearth patterns and anomalies that an insurer may not have even thought to look for, and may have otherwise escaped notice. For example, an average claimant with a particular condition may have five consultations before a scheduled surgery, but AI could uncover an unnoticed group with a completely different, but shared, treatment pattern. AI can also refer to robotics and other applications that go beyond simply analyzing data.
Also, while many insurers have access to vast amounts of data, most of this information comes in structured formats, such as an organization by codes for diagnosis or treatment. Few insurers have large enough unstructured data to enable AI to teach itself and fully mimic human decision-making. AI technologies are still being tested within the life insurance sector, and applications may also be constrained by complex variables across multiple product lines, regulatory concerns, and privacy considerations.
In summary, despite long-term underinvestment, claims functions should demand a higher level of innovation than other areas of the insurance ecosystem given the sensitive nature of the claims experience. The claimant is vulnerable, and swift, supportive, and decisive action is required. Those who invest in improving the claims journey — instead of viewing disruptive technologies as a threat — will increase customer satisfaction, lower claims processing costs, and carve out a durable advantage now and in the future.
3b6df65b-f107-43ac-9b32-d8a308104d95.png?sfvrsn=8a342d60_3)