10 ways AI & Machine Learning can increase insurance revenue.

Last week, Advancing Analytics attended BIBA 2022. It was a great opportunity to catch up with the best and the brightest in the Insurance and Brokerage business. There were fantastic talks on the direction of the industry, and one are which is getting a lot of attention it the application of AI and Machine Learning to Insurance. The Insurance sector is ready for disruption, and it will be AI that causes the disruption. The only question is, will you be the disrupter, or get disrupted?

You may be wondering where you should begin with Machine Learning and what are some of the ways in which Machine Learning and AI can drive profits and better customer engagement? For personal lines cover, approaches generally breakdown in to revenue protection or customer engagement. (Not sure what the difference is between Machine Learning and AI? Take a look at this blog for more details)

AI for Revenue Protection

As the cost of living increases, everyone is looking at how they save as much as possible. The impact of that of this forces customers to evaluate their level of cover and as the cost of living increases, so does the risk of customers cancelling/defaulting on payments. Artificial intelligence can be used to protect revenue in a number of different ways. In this blog we will look at 10 ways in which AI can be used to protect revenue.

1. Lapse identification. – Identifying those policies which are likely to lapse on a policy or default on a payment. Early intervention can have a huge impact. Reminding a customer why they use your services and what you can offer, greatly improves retention.

2. Default identification  - – Identifying those policies which are likely to default on a payment Where a customer is showing signs of potential defaults, early identification ensures that you are aware of the impact before it occurs and you have the opportunity to intervene before and debt collections are escalated. Machine learning is only as good as the data you hold, and to make the best model possible you need to look at other factors. Open data, sanctions, ccjs etc. When combined with known customer trends and data a classification model will yield great results.  

3. Customer assistants – How much of your call centre work flow could be optimised/automated using chat bots? Bots will not replace the need for call handlers, but if you can reduce the average workload by just 30 minutes a day per handler (~6.25%), assuming you have 50 handlers, that is a total saving of approximately 300,000 minutes or 40 weeks! Increasing this to 20% could save the salaries of 5 call handlers and increase customer experience. Customer assistants can also speed up manual back-office activities such as document reading and document understanding.

4. Claims predictions – Which claims have a high propensity for significant payout? When modelling revenue there are many factors which impact profitability. Claims is a volatile are for predictability. Understanding wider factors (weather, the political landscape) which might put properties/vehicles at risk. The shortage in car parts, building materials and second hand cars is forcing the value of claims to increase. Understanding the potential cost of a claim as early as possible or before a claim is made can allow for increased trust when deciding to move to re-insurance.

5. 3rd party claims predictions – Which policies are most likely to end up going to a 3rd party claim? Who is at fault based on all the collected information?

6. Claims Fraud propensity – Which claims are suspected as Fraud? Fraud is always a challenge, and seldom does one approach provide enough assurance that a claim is fraud. An ensemble of methods will also have the highest chance of success when combined with novel approaches such as network graph traversal. We hope fraud is a minority in you business, less than 1%. You need to be careful with low numbers, as we see what is known as a class imbalance which tends towards models which claim to offer high levels of prediction.

7. Demand forecasting – Are we seeing an increase/decrease in demand? Is a product which used to sell very well slowly declining? Is our cover portfolio still competitive? When looking at the numbers daily it is hard so see what is really happening. Personal lines can be very seasonal, especially where cover is aimed at a student market. Understanding and anticipating demand give you the ability to confidently predict revenue or know when to explore alternative options.

8. Revenue Forecasting – Are we seeing an increase/decrease in revenue? Are our prices still competitive? Are we spending too much on claims? There is many factors which influence revenue. To build a successful revenue model, you need to look much wider than just how much revenue is currently generated. Revenue is impacted by everyone of the approaches discussed in this blog. Reduction in customer churn, reduction in claims fraud, better control of policy lapse and defaults will all drive your bottom line. Understanding demand and the levers which control revenue will transform the way you do business.

9. Assessor assistants – Think Hal 9000 but for Insurance! Using deep learning to process images of damage. How much time is spent inspecting images of motor/household damage? How sure are you that the images send of structural damage are for the home which is covered? Can you estimate the total claim cost compared to the value from a series of images. Using deep learning we can achieve all this and more. You might think that is beyond what you can do today. In recent years image processing has been commoditised, and image classification can be achieved very quickly.

10. Advanced insurance – Taking a photo of the property can identify high risk features, flat roof, trees near the property. We are seeing more and more that finance is becoming personal. Personal cover plans look to be the future. Rather than a policy aimed around the way the average consumer behaves, lets look at how you behave. If you are a low risk customer, we can reduce premiums or create a plan that is designed to meet you needed.

At Advancing Analytics we understand Insurance. We are experts in Insurance and Financial services having helped customers in personal, commercial and brokerage. Our team of experts are on hand to help you increase revenue with AI and Machine Learning. Are you new to Machine Learning and not sure how to get started? Then take a look at our Managed Machine Learning Service. We can take away the pain of hiring a team, while delivering value early with our Insurance focused solution accelerators.

Click below to schedule a call with one of our Financial Services experts.

Looking to understand how AI can be used to increase customer experience? Then you will want to check out our next blog.