As insurance companies well know, automating claims and underwriting processes typically involves a trade-off between speed and accuracy. But advances in artificial intelligence-based automation platforms continue to change the equation, helping insurers process claims faster and reduce underwriting leakage.
We’ve written about the speed vs. accuracy dilemma before and pointed out how intelligent intake platforms can eliminate some 70% of manual document handling, speeding up processing for both claims and underwriting submissions.
Now AI automation models are getting even better. Independent testing shows recent improvements further reduce time spent on task by 33% while reducing errors by 20%.
Both of those numbers are significant. The first speaks to increased productivity of your claims and underwriting workforce while the 20% figure directly relates to reducing costs. Errors can cost insurers in a number of ways, including lost business because of quotes that are too high. Perhaps worse is erroneous data being incorporated into risk models, potentially skewing hundreds or thousands of policies.
Let’s dive into how upgrades in the latest iteration of the Indico Unstructured Data Platform, Indico 6, increase both speed and accuracy.
A human-in-the-loop assist
In addition to enhancements to make filtering and sorting easier, a big enhancement in Indico 6 has to do with making human-in-the-loop corrections to existing models. Human-in-the-loop is a crucial capability that enhances a model’s effectiveness. Essentially, it involves making corrections that get incorporated into a model, to consistently increase its accuracy over time.
Indico 6 includes new calls to action that help users home in on the data the model is least sure about. It helps user ensure their time is spent where they can add the most value, as opposed to areas where the model is already doing a good job.
The idea is to highlight areas where the model has low confidence that it’s selecting the right data from a given document. If something doesn’t conform to a specific, predefined business rule, for example, the user may get a notification to take a look and see what needs addressing.
Such issues are more likely to crop up in complex submissions that contain multiple document types, such as an underwriting submission that contains an email message body, loss run report, older policies, a broker report, and the like. Giving users an assist in where to make improvements in such submissions means they can process all future ones that much faster.
Related content: AI loves data and insurance has lots of it; the possibilities are endless
Microsoft Exchange integration
Another way to speed up automated processing times is by making integration with other systems easier. Toward that end, Indico 6 now enables easier integration with Microsoft Exchange-based email systems.
We’ve heard from customers who experienced throughput problems when using other tools to pull emails from Exchange-based email systems and then feed the emails to Indico. Indico has always been able to handle a large amount of throughput, so we took this gating factor away by enabling users to configure Indico to directly connect to Exchange.
Users can have one or many email boxes associated with a given workflow, such as one dedicated to claims and another to underwriting. With just a couple of pieces of information about the inbox, users can set up the connection on their own, without involving IT.
Related content: How generative AI is revolutionizing insurance: A conversation with Indico’s CEO and Microsoft’s AI expert
These are just a couple of the improvements you’ll find in Indico 6, along with improved handling of data in table formats, and enhanced labeling capabilities. To get the full picture, consider signing up for a free demo. Or, if you have any specific questions, just contact us.