Trusting in algorithms

Ever since Edmund Halley, the man of the famed comet, calculated the first life expectation tables in 1693, data have been the fuel of insurers. Three hundred years later, data remain essential to insurance, but the swelling seas of information are also being exploited by many other companies as well.

Internet giants have refined the art of collecting petabytes of data and applying mathematics and increasingly sophisticated algorithms to extract hidden insights, meanings and probabilities. Such knowledge is being used to develop advanced new technologies.

The internet giants are typically looking to predict your next purchase, or a new contact in your social network, whereas the traditional strength of insurance is to predict risk: what is the likelihood a client will have a loss due to a wildfire, cyber attack, construction accident or supply chain disruption? 

New data sources and technologies, such as machine learning, present exciting new directions for predictive services. The end goal is to develop new solutions and products around these new advances, moving insurance from “react and respond” to “predict and prevent”. 

Predictive services and machine learning are two technologies within the field of artificial intelligence expected to have the biggest impact on industries in the coming years, according to AGCS’s recent research.

Assessing 25 trends in six different categories including Socio Economy,  IT Infrastructure and Cities & Mobility, the AGCS Trend Compass is a new strategic tool to help companies navigate the complex, developing landscape of opportunities and risks over the next five to 15 years. More than 200 AGCS managers and risk experts identified predictive services as a trend that will have a very high impact on client companies.

TrendOne, a market leader in innovation and research, then assessed the speed at which these trends will be adopted. They estimated that it will require the immediate activation of resources for rapid adoption within the next five years.

Using remote sensors and cloud-based software, predictive maintenance software is capable of assessing machine and industrial assets in real time. By monitoring acoustics, vibrations and temperature, the sensors generate feedback that can recognize deviation of performance patterns and potential part damage or failure, in which case the operators are contacted.

Thomas Gellermann, Expert of Allianz Center of Technology, AGCS: notes that, “More and more industrial and engineering companies are introducing predictive maintenance systems in their maintenance strategy.”

Gellermann says early detection or prediction of cracks, for example, can prevent the affected component from breaking, preventing damage to adjacent parts or even the failure of the entire machine. “Predictive maintenance is a powerful technology to mitigate physical damage and costly business interruption as a result of standstill of machinery,” he notes. 

Another examples of how Allianz is applying such technologies, is a partnership between AGCS and Praedicat, an InsurTech company that uses Big Data to modeling to predict with increased accuracy the key catastrophe liability risks ahead.

It can take years, even decades, for the widespread human and environmental exposure to play out. DDT was once hailed as an efficient pesticide; asbestos was used as far back as the 18th century; and plastic micro-beads were considered good exfoliating agents in cosmetic products – until they were found to have a huge impact on the food chain. 

Praedicat’s solution scans, analyzes and synthesizes data from millions of peer-reviewed scientific journals to identify product risk, helping scientists reach new levels of analysis in a much faster and efficient way.

Under the partnership, the underwriting processes of AGCS are integrated with Praedicat’s predictive modeling approach. The combination means AGCS liability underwriters now have access to solid data that allows them to better assess potential ‘unknown’ liability risks for industries or even individual companies with higher confidence. 

With a few clicks in the digital tool, they can learn if a specific chemical substance such as for example Bisphenol A replacements is becoming more critical over time as adverse health effects or even litigation are reported in public sources.

“Forward-looking analysis will transform insurance underwriting,” says Hartmut Mai, Chief Underwriting Officer and Board Member, AGCS. “Through predictive modelling, we hope to change the core role of underwriters, freeing them up from the daily paper grind and empowering them to be data scientists.” 

For sure, there is no shortage of exciting developments in AI and sensor technology for around insurance  – many new types of insurance and mitigation services will emerge that are data-driven, predictive and preventative.

“However the real challenge is to determine which of the new breakthroughs make sense for our customers and us as their partners in managing risk”, says Larsen. “The transition from managing risk after the event to predicting and partnering with clients will require not only algorithms but also our risk expertise and a laser focus on our customers’ needs.”