How Automated Machine Learning is reshaping insurance pricing

Artificial intelligence is changing our lives in many ways. Insurers can use data and algorithms to predict the behaviour of their customers. This also offers new opportunities for underwriting and pricing.

3 minutes read

Published 02/14/2022

How Automated Machine Learning is reshaping insurance pricing

Experts
Related Topics
Share
LinkedIn

Facebook

Whatsapp

E-Mail

Copy Link
Sharing

Massimo Cavadini, Head of Actuarial Consulting & Data Analytics at Munich Re, explains how automated machine learning (AutoML) puts the pricing expert at the heart of the decision-making process and how clients benefit from this.

Insurers want to increase their profitability. What role does pricing play here?

In the context of the current market situation with new risks and continuing low interest rates, companies are forced to focus on technical results to improve profitability. This means that there is a need for a better risk selection and an improved pricing process. This is made possible by the vast amount of structured and unstructured data now available to insurance companies. Fed into machine learning algorithms which are run using huge computational power provided by cloud computing, it will boost underwriting and pricing sophistication to a new level.

How is this achieved and what impact do these developments have on the pricing and the underwriting cycle?

The previous industry standards for prediction as a basis for risk, behaviour and market pricing were generalised additive models (GAMs) which represented the industry standard due to their explicability and auditability. These have been matched by new machine learning techniques, including a so-called tree-based approach to predicting future costs and customer behaviour. This has improved prediction quality, although their usage is limited for rate-setting due to their lack of transparency.

In the last two years, the new field of Automated Machine Learning (AutoML), called also explainable AI, offers a compromise encompassing the advantages of machine learning and the explicability of GAMs.

AutoML has started to gain traction in both life and non-life insurance in both the pricing and general data science areas of the business. It shows great promise in providing explainable and implementable results and allowing for greater access to the efficiencies and model predictiveness of machine learning.

Can you give us an example?

Here in Global Consulting, the consulting arm of Munich Re, we work closely with our primary insurance clients helping them growing profitable. Quite recently, we leveraged the full capabilities of AutoML to help a digital insurer estimate future costs and customer behaviour, running 45 thousand models to identify the best one. The result was reflected in a projected Loss Ratio improvement of 4%.

What does this mean for the role of the pricing actuary?

AutoML represents a paradigm shift as it commoditises predictions, allowing the pricing actuary to focus on the decision-making process and implementation. Once the data are collected and cleansed, the pricing practitioners will run different models to predict the future costs and customer behaviour.

This allows a single analyst to run hundreds of models and interactions at the same time, testing different hypothesis within a short time frame. The practitioner in turn can focus on decision-making, where experience an domain knowledge plays a relevant role.

Domain knowledge plays a paramount role in this decision-making. More than the capabilities to model, which can be outsourced and replaced by a machine, the practitioner must demonstrate sound domain expertise in order to make use of the prediction process.

Can you tell us how this will be used in the future?

We expect AutoML to get more relevant in the near future as an integral part of the pricing and underwriting cycle. The insurance pricing software landscape is evolving quickly. Munich Re is continuously searching for market solutions to meet the diverse needs of our clients.

For example we recently teamed up with a French start-up, Akur8, which is one the most interesting players in this space. Leveraging on that, we are now able to provide our clients with the best combination of risk transfer, domain knowledge and technology.