According to the From art to science: The future of underwriting in commercial P&C insurance report from McKinsey, “underwriting excellence remains paramount to company performance” and “operating results—more than capital leverage or investment returns—has the greatest impact on overall financial performance.” Additionally, while comparing the top and bottom quintiles of P&C companies, McKinsey found that the biggest differentiating factor was, far and away, the underwriting.
Loss runs are one of the most critical elements to quality underwriting—but are a document format that has been historically challenging (for both humans and machines) to understand, due to the wide variability in formatting, nested tables, and graphical elements. While humans can comprehend that certain data points need to be interpreted together to make an assessment on a loss run (e.g.: the name of the claimant and the claim value, which can be positioned in different parts of the report), legacy technology (such as open source NLP models) is simply not designed in this way.
Due to the vast array of limitations inherent with legacy automation technology, Applica built an entirely new way to accelerate business operations. Our progessive AI-powered automation solution leverages our own generative language language models to understand complex and variable document types such as loss runs. Thus, Applica is able to give meta attention to them, which enables a machine to understand data points that only make sense if taken in context with other data points within the document.
Because of this meta attention feature in Applica’s solution, insurers can now process loss runs in a totally automated manner, finally retiring the laborious process of employees reviewing each document individually. And with a staggering 80% of insurance information currently buried in disparate paper stacks and digital repositories, automating the loss run process enables users to extract a vast amount of customer information. This in turn allows for the creation of data lakes that can be leveraged to make more informed underwriting decisions.
Seems too good to be true? Let’s take a look at some visual examples.