Three ways artificial intelligence will transform commercial insurance

If you want business insurance in 2017, chances are you’ll be dealing with a bot not a broker. An artificially intelligent system will scrape all of your valuable information from the web, create a holistic representation of your business, and decide if you are eligible or not - and at what price - within a matter of seconds.

More information exists on the web now than ever before. Every two years, we create ten times more data than has ever existed in human history [1]. By 2020, technology conglomerate Cisco, predicts we’ll be swimming in 2.3 zettabytes of web data [2], and strategists at the UK Ministry of Defence believe that by then, 90% of what happens in the physical world will be described in unstructured web text.

In this post, we are going to briefly explain three ways that AI can help insurers utilise this data to enhance their operations;

1. AI gives insurers access to an increasing number of observations about risk

2. Improved risk pricing

3. Better risk selections


Insurers can now use AI to gather web data and turn it into intelligence. Actuaries and analysts have been slicing and dicing their internal data for decades, but they have never had access to so much of it from third parties before. AI enables insurers to gather and synthesise unstructured and structured data from disparate sources all over the web, and weave it together with their existing intelligence. 



Using AI, insurers can generate a more factual, robust representation of commercial risk that is built from more than just historical claims data. For commercial insurance, an industry that is heavily dependent on statistical analysis and predictive modeling, this is a pretty big deal.


Commercial underwriting traditionally relies on the analytical expertise of humans and their ability to manipulate historical datasets. Emerging risks such as cyber attacks, where claims data is immature or sparse, are difficult to underwrite. Exposure is hard to predict and difficult to price.

This makes it incredibly tricky for insurers to provide intelligent and competitive pricing to the masses of business owners who want cyber coverage, further illustrating the point that these days, a lack of data usually leads to a lot of missed opportunities.

In these cases - products converge towards the norm and exclude the profitable opportunities that exist at the margins or in niches. Machine learning methods such as deep learning solve this problem.

Deep learning uses multi-layered neural networks to learn and can apply complex mathematical calculations to data over and over again, learning from each interaction. This is especially useful for identifying significant patterns that are difficult to spot in vast amounts of data and help commercial underwriters to extract new insights incredibly quickly.

AI helps by introducing new data from the web - such as a complete history of reported and unreported data breaches, and the characteristics of companies incurring them. It then analyses this data at lightning speed and rapidly predicts the probability of loss.

Ultimately this means insurers are able to underwrite new risks in emerging markets faster than ever, leading to new sources of revenue, and more competitive pricing.


With regard to Cyber risk, instead of solely focusing on the revenue and sector of a company to predict loss frequency, underwriters can consider factors like media profile, reputation, and political stance.

When a large business receives negative media coverage, does this increase the likelihood that they will experience a cyber breach?

The use of web data combined with machine learning makes it possible for insurers to answer questions like this. Equally, information that commercial insurers used to rely on customers to provide can now be pre-populated in seconds using services like Google Maps, OpenStreetMap and LinkedIn.

Having access to all of this insight makes it increasingly easy for insurers to identify good risks, and allow them to focus resources on creating profitable products in these areas.

There is a possibility that soon, AI will be able to create a true representation of risk. It’s not really a question of if commercial insurers will embrace the capabilities of AI, it’s a question of who will do it better, first.