Take the holistic insurance model a step further: Data lakes integrated with artificial intelligence

In my previous blog, I covered how insurance analysis has moved on from traditional methods to holistic models. These models allow insurers to better assess risks and losses by being able to view their entire portfolio in a non-siloed way. This allows insurers to make better use of the data they hold and provide more proactive support to their customers.

Insurers cannot move to a proactive model without getting their own house in order to better leverage the data they hold. So, the question is how do insurers reach this point and what kind of customer will benefit from it once it is in place?

Insurers can go and jump in the lake if they don’t fix their data

Insurers have a number of technologies to choose from to automate their processes and ensure better collaboration with their customers. Artificial Intelligence (AI) is one such technology. Research from McKinsey found that AI implementations could increase productivity in insurance processes and reduce operational expenses by up to 40% by 2030. Therefore, by investing in technology such as advanced data analytics and AI, insurers can identify fraud more easily, reduce their costs, and automate claims processes to give time back to claims handlers to actually speak to their customers.

In order to leverage AI properly, insurers have to pay attention to the quality of the data that they are using. Data cleaning and readiness need to be priorities in order to avoid false positives or biases creeping in. The amount of up to date and relevant data that an insurer has at their disposal is also critical, particularly as it relates to informing the real-time decision making that AI promises to unlock.

Having all of this data means nothing, unless insurers can properly aggregate it in order to find patterns. One way insurers can do this is through data lakes. These represent large storage repositories which hold vast amounts of raw data until it is needed. Insurers can then use AI models which can transform the data into insights that are actionable and drive future decisions.

Insurers must also have a core platform that enables them to integrate third-party data sets which can be leveraged to add greater context and up to date information about what is happening on the ground, so they can make more informed decisions. This data could cover everything from weather patterns, water levels, the location of wildfires, the frequency of cyberattacks, or the transport and location of goods. In order for insurers to make this work, their systems and workflows need to be able to incorporate information from both internal and external data sources that drive better outcomes for the insurer and the insured.

With the shift to cloud-based core systems that are built to integrate and share data and information, there is an opportunity for insurers to pool data at a market or regional level (subject to data privacy regulations) to drive greater efficiency and build more comprehensive models. 

Whilst this is not common today, as risk becomes more complex insurers will see more value from pooling their data with others instead of going it alone. You only have to look at the size of the data sets being used to drive AI innovation in large language models, for example, to understand that the data held by one insurer is probably not enough to create real value. 

Equally, biases in a single set of data will almost certainly exist. For example, say an insurer covers five out of 200 properties on the same street in a high-flood-risk postcode. Let’s assume these properties are located higher above sea level than those surrounding them, and so historically damages have been less. Without a fuller data set, the insurer might build an AI model with a skewed vision of reality. 

This is obviously an extreme example, but it highlights the problem when bad or incomplete data is used to build AI models. The term ‘hallucinations’ is used to refer to confident but misguided responses made by AI. This really highlights the need for a diverse data set that can be cross referenced, validated, and complete. 

What does this mean in the real world?

Once insurers have put the right models and solutions in place, it should be possible for them to be much more proactive in how they mitigate risk and support their customers.

How does this look in practice? Imagine that there is a severe storm forecast to hit a particular area. Ideally, insurers will have the rainfall forecast, storm surge, and wind speed data integrated into their models to enable them to understand the degree to which different properties and businesses in their portfolio will be affected. This insight can then be translated into action; warnings sent via SMS services to insureds letting them know that there is a risk of weather damage and sharing tips on minimising it, such as moving valuable items or furniture from the ground floor; sandbags automatically ordered and shipped to properties to help bolster flood defences; wooden panels sent to block windows; plane or train tickets automatically booked to a predetermined location so that insureds can evacuate to a safe place whilst the storm blows over. For business customers, this might look like goods being rerouted or transported to different storage facilities that will not be affected by the storm, or hotel space being automatically booked as a back-up office to allow the business to continue to run if office space is taken completely offline.

By being more proactive and offering personalised support, insurance customers have time to protect not just the things that they hold dear, but can get themselves out of harm's way as well. That text message or email might not just contain information on what to do at the property; it could also include a hotel booking confirmation which is pet friendly to enable the insureds to take their furry or feathered loved ones along with them to safety.   

Don’t ask me, just sort it out

Regardless of whether it is for businesses or private consumers, it is now possible for insurers to take a more active and consultative role in the lives of their customers. This allows the insurer to become a real partner, and as long as this is all seamless and largely automatic, requiring little to no input from the insureds themselves, provides a meaningful value added service. Ultimately, this is where the real value lies and what will help insurers move from a transactional relationship with their customers to trusted partners. Doing this across an insurer’s customer base would have previously been unthinkable, but using technologies available to us today it is now entirely possible. After all, when people are going through a traumatic event, taking care of things on behalf of the insured to make it a little easier is just the right thing to do.

Whether insurers can deliver on this promise or not depends on whether they can do the unexciting but critical work upon which the innovation can take place and true societal value can be realised. More on that in my next blog.