How do insurance companies work with NLP to detect fraud?

Insurance companies are a largely heavily regulated paper industry, so claims even the retail side where you’re applying for insurance, everything is text-based and we’d love to believe that you go online and you click a few things, fill out a few forms, and you are insured. 

Most insurance companies, other than maybe some small niche super new insurers, who are in this like the new innovative world still want all this stuff back by paper.

First, you have the application. You want to ensure that your applicants are who they say they have the history that they say they are, their history comes with their license or their prior insurance.

What you have is a lot of insurance companies have systems or people in places that will do a bit of applicant research, so while you’re applying to their back office look through many signals, including any history they may have on you.

The other way that they use AI and NLP is on the claim side, so doing a similar thing and looking for signals that may state, why you have such a claim, it’s insurance so it can get messy, imagine somebody going through a divorce and they must pay out to the other party.

It’s possible to get a bunch of insurance money and pay out the other party using that, so maybe you falsify a claim on your car getting stolen or wrecked or your house or something, something along those lines.


Let’s cut through the jargon, myths and nebulous world of data, machine learning and AI. Each week we’ll be unpacking topics related to the world of data and AI with the awarding winning founders of 1000ML. Whether you’re in the data world already or looking to learn more about it, this podcast is for you.