All-or-Nothing Automation? Incremental Change Is a Winning Proposition

By Abhishek Lall  |  November 15, 2019

Until recent years, implementing automation improvements in insurance processes seemed a futuristic dream available only to carriers with the biggest IT and R&D budgets. Today, thanks to economies of scale brought to bear by open-source data processing and widely available cloud storage, insurers have better access to solutions that can make the ambitious goal of automation more realistic.

However, the prospect of end-to-end claims automation feels like the wrong approach for a number of reasons. First of all, during moments of devastating loss, adjusters provide caring personal service to claimants and thoughtful oversight to the claim process, which is at the heart of what distinguishes the best insurers. 

In addition, the idea of yanking out all manual claim processes and replacing them with automated systems feels like an implementation nightmare that few insurers would be willing to take on. The cost of such a drastic move would certainly be prohibitive, and the disruption to workflow as well as the time to work out all the kinks could leave both claims staff and customers dissatisfied.

The good news is that automating claims isn’t an all-or-nothing commitment. In fact, the best way to consider bringing in automation is to determine first which discrete activities that are now handled manually would make the most sense to switch to computerized processes.  

A good example is medical records. Most of these claims aren’t submitted electronically, and more than half of third-party medical bills don’t make their way into claims databases. With no uniformity, this unstructured data must be reviewed manually. But medical records aren’t easy to scan visually. Each physician’s office puts the invoice number, ICD-10 code, doctor’s notes, and so forth in a different place on the form. Furthermore, looking up medical codes is extremely time-consuming when hundreds of pages of records are being evaluated.  

To sort through up to a thousand pages of treatment details that an attorney may offer would take an individual working full-time day on end on just that case. No adjuster has that much time to focus on a single claim. 

Yet, when an attorney puts an offer on the table, the insurer must respond in a timely fashion—putting even more pressure on the claims representative to quickly digest all the treatments, medications, and possible surgeries being requested. Without tools to help dig into the true costs, insurers are reaching decisions to accept or reject offers without critical insight.

A better approach would be to turn to an AI-based automated system that’s trained to “read” medical records, digest the information, and draw certain conclusions about potential claim severity. With this insight, the insurer would be in a much better position to make an informed response. 

In the example of medical records, a machine learning system could be trained to look for relevant family history (for example, Alzheimer’s, heart disease, diabetes). The system could automatically review all the symptoms presented and consider what the diagnostics show. Are there indicators of whiplash or other soft tissue damage? The system would also consider the medications being prescribed. High-cost, addictive opioids would be flagged for special attention. An automated system could quickly find those, whereas manual processes might too easily miss them. In addition, the system would check demographics such as age, whether the claimant smoked, and more.  

Adding automation to the medical records review would turn a tedious, time-consuming process that provides limited information into a fast, automated activity that offers real insight into potential claim costs and recovery times.

With these AI insights, the claim could also be better triaged. Right from the start, the most complicated high-severity claims could be assigned to adjusters with the strongest experience in complex bodily injury claims.

Evaluating medical record information is just one example of where automation can make significant improvements in the insurance claims operation. Robotic process automation (RPA), machine learning, and AI can be embedded in processes throughout the claim life cycle—from triaging through settlement. With tedious manual processes off their plates, adjusters would have more time to manage claim tasks that require their experienced oversight.   

Verisk is at the forefront of bringing automation to insurance claims with a range of predictive analytic, AI, and other automation solutions that save time and improve results. To find out more, contact me at Abhishek.Lall@verisk.com or +1-816-352-9370.


Abhishek Lall

Abhishek Lall is vice president of analytics and leads the injury analytics team at Verisk. He has more than ten years of progressively responsible experience providing leadership, direction, and functional expertise to design, develop, and deliver advanced analytics on complex initiatives. Before joining Verisk, Abhishek led a team of data scientists at Travelers responsible for developing predictive modeling solutions for fraud detection, risk control, and other claims initiatives. He is also the author of seven U.S. patents based on advanced analytics.