What is Machine Learning and How Can it Help Insurers?

What is machine learning?

Machine Learning is very much in the spotlight. It underpins many of the day-to-day activities we do, such as the Google search engine, Amazon’s “Recommended for you” and Uber’s estimated arrival times. This is the tip of the iceberg and many believe that it will transform our lives – from driverless cars to the ability to talk to anyone in any language.


How can machine learning help insurers?

This technology is also present within the insurance industry with initial implementations focused on calculating prices and combatting fraud. At LV=, we’ve been working with machine learning for two years and have a specialist team of eight data scientists and engineers. As Director of Pricing and Data Science, it’s my job to make sure we’re using machine learning as effectively as possible. Here are just a few bits of insight from me, based on my experience over the last two years:.


1. It’s not about pricing

We love machine learning in pricing and how it can be used to improve risk selection, but we have found the really exciting projects are in the wider business. Improving the claims process is a great example. Just look up the news on Lemonade in the U.S. paying a claim in 3 seconds through machine learning. At LV=, we have over 200 machine learning use-cases we are working through, ranging from counter-fraud to HR to IT operations .


2. It doesn’t have to be expensive

You can spend – and many have – millions of pounds on machine learning projects, but that doesn’t need to be the case. Many models are open-source, which means that they are free. For example, type “TensorFlow” into Google and you can access some of the most advanced “Deep Learning” code for nothing. Processing data can be done on a “pay on use” basis in the Cloud. From our experience, the main cost will be the changes to your business systems to use the machine learning, but we have used manual workarounds here just to get it up and running and see the benefits.


3. You don’t need huge amounts of data

The more data the merrier, but a spreadsheets-worth can be enough for many projects. Some machine learning models are even designed to be more effective on smaller amounts of data.


4. The outcome is business transformation

A data scientist can build a supremely intelligent model, but unless it does something that improves profit or customer experience it is has little value to us. We make sure someone who understands the business leads every machine learning project .


Source: Hugh Kenyon, Director of Pricing and Data Science, LV= General Insurance