The modern actuarial office: Why AI and speed will generate sales in the future


Some say that the insurance industry is a long, quiet river on which stately steamships cruise. Others say it is a shark tank where only the strongest survive. Which is right? 

The answer is both. The insurance market is clearly mature, with a limited scope of action for individual players. There is, however, no question that merciless predatory competition is taking place, probably precisely because of this saturation. It is the perfect recipe for a ruinous price war, even though this is something that insurers really cannot afford in the long term.

Right in the middle of this tension are actuaries. They have to model tariffs that are both attractive for the customer and profitable for the insurer. This has always been the case, but now the industry is accelerating.

Insurance products today have to be solid and fast. The internet, and particularly comparison sites, have brought total transparency to the market. This has fundamentally changed customer behaviour. Customers now look for the cheapest tariff, and this is driving prices down. It‘s a bit like buying petrol. If Shell’s petrol is three cents cheaper in the evening, Aral’s regular customers won't have to wait long for a reduction. This means that insurance companies are having to become increasingly agile in their pricing. Actuaries must be able to adjust tariffs rapidly, as well as develop new ones and convert them into products.

Agile product design, outdated processes

Unfortunately, actuarial systems have often grown organically over a long period. They rely on processes and structures developed over time to meet particular needs. Many are manual and involve system breaks. This results in a high risk of errors, and also means that an end-to-end run takes a long time and is not easily iterated. 

What’s worse, as a rule, is that actuarial systems use standard modelling procedures for tariffs. They use the same data and variables year in and year out. This leads to huge challenges when real-time modelling is required. Tariffs are often still stored in programmed calculation kernels. Without IT support, no adjustments are possible. A tariff change or the implementation of a new tariff takes a very long time. This is the polar opposite of the fast, agile and flexible service that customers, the market and sales channels are demanding.

As a rule, is that #actuarial systems use standard #modelling procedures for tariffs. They use the same data and variables year in and year out. This leads to huge challenges when real-time modelling is required. Click To Tweet

Is the answer artificial intelligence? You might think so if you listen to the hype and watch trends in the industry. Fukoko Mutual Life in Japan, for example, has already replaced many employees with AI. I don’t think anyone would go that far when it comes to actuaries. One thing is clear, however: Actuaries have to change from being tariff and model administrators to innovation drivers in insurance. Technology can and must help here, and AI and machine learning are ideal for this. 

Modelling and machine learning: A dream couple

If you take a look at the entire process from collecting data to deploying tariffs, there are plenty of modernisation options for each step. It is possible to transform actuarial processes from end to end. For example, machine learning methods are available on both visual and programming interfaces, making them easily accessible and usable directly by actuaries. The combination of machine learning and modelling could easily become a key actuarial competence in the future.

Changing actuarial and insurance processes should not be an earthquake that leaves nothing untouched. There are numerous well-functioning processes in insurance companies. New solutions can easily be wrapped around them, ensuring that well-functioning elements can be integrated and actuarial excellence increased. 

The fact remains that actuaries, with their experience, sense of proportion and creativity, will remain the backbone of any insurance company. They do, however, need new tools so they can act quickly and shorten the time to market. There is no danger of algorithms replacing actuaries; instead, they will use technology to enhance and extend their work.

What’s next:

AI and algorithms are only part of the solution. The best tariff is useless if it doesn't get to the customer quickly. You can find out how this works in my next blog post.


First published on Mehr Wissen blog: Das moderne Aktuariat: Warum KI und Tempo künftig Umsatz bringt 

About Author

Michael Rabin

Michael Rabin assists insurance companys on their way towards digitalization, big data analytics and IoT. Prior to this, he worked as a technical account manager in this segment. He began his career with a German insurer (including businesscontrolling of property / casualty insurance and strategic bankassurance) and is therefore familiar in the field of classic insurance. Michael Rabin unterstützt Sie als SAS Account Advisor für Versicherungsunternehmen auf ihrem Weg in Richtung Digitalisierung, Big Data Analytics und künstlicher Intelligenz. Zuvor war er als Technical Account Manager in diesem Segment tätig. Seine Laufbahn begann er bei einem deutschen Versicherer (u.a. in der Geschäftssteuerung Schaden-/Unfallversicherung und im strategischen Bankenvertrieb) und ist daher auch im Bereich der klassischen Versicherungsthemen zuhause.


  1. Very interesting topic and certainly one that has occupied my thoughts for some time now. While insurance companies have tons of data which ends up not being used, there is big potential to implement ML algorithms for different purposes, mainly recommendation systems or customer analysis. However, I don't see its application in the actuarial field simply because there are strict regulations by law about how tariffs are allowed to be calculated and changed. For P&C the regulations could be slightly more relaxed and thus ML could possible be applied instead of the standard GLMs, but for life & health, it might take decades of regulatory changes for ML to be allowed to be applied within the tariff calculation.

  2. Thank you very much for your comment. Indeed, it is true that machine learning algorithms may not yet be used everywhere as a tariff model instead of a GLM.
    Nevertheless, this is an exciting new field for actuaries. I experience daily that actuaries have started to deal with modern modeling techniques or want to start doing so soon. Actuaries, so to speak the top dogs among the data scientists in insurance companies, see what advantages this can already have.

    On the one hand there is the approach to ML and the understanding of what it is and how it works. And on the other hand, the desire for innovation, from the point of view of technology and methodology, but also from a professional point of view with a view to productive tariff models.

    Even though it is clear to all involved that we will see GLMs in production for a while yet, ML (best combined with a visual intuitive front-end) can already support actuaries today in exploring known and unknown data quickly, discovering correlations and quickly testing model ideas for their viability in the sense of prototyping.

    The most important advantage in my opinion, however, is the possibility to "challenge" GLMs with machine learning models and to see what machine learning models do differently, for example in the selection of variables, in order to improve already productive GLM tariff models with the newly gained insights.

    By the way: Customer analysis and recommendation systems in marketing are closer to the actuarial profession than one might think at first glance. Boards of directors and actuaries are increasingly interested in dynamic pricing, where the aim is to provide a potential new customer with an individual price offer in real time at the point of sale. In addition to the pure tariff model, business rules and advanced models are used to include, for example, customer behavior, competitive prices, conversion or discount models in the pricing for the offer.

    So at the end of the day, this is a decision based on data and analytical means to create the best possible value. The same is true in marketing, when insurance companies implement topics such as Next best Action, Offer or Question. The implementation is even possible with the same technological base and the same software and has strong synergy potential for insurance companies.

Leave A Reply

Back to Top