AI Case Studies: Smart Homes, Smarter Insurance

What if your home could prevent insurance claims before they happen?

As smart home technology becomes increasingly accessible, traditional home insurance models are changing - and actuaries can be at the forefront of this revolution.

In our new series, AI Case Studies, we share hypothetical examples to illustrate how AI may transform insurance processes in real-world scenarios and how actuaries can use AI to help benefit the community. This first case study in the series focuses on home sensors and their impacts on general insurance products. 

Home sensors

XYZ underwrites home and contents insurance. XYZ's risk assessment and pricing has traditionally been based on variables commonly associated with this class of business (for example location, building material, claim history). XYZ is investigating the use of sensors and smart home devices to mitigate risk and reduce claims cost through early warning.

XYZ recognises the opportunity to leverage the data available from more smart devices being installed by homeowners. This may lead to:

• Better risk assessment and differentiation through enhancements to underwriting and pricing processes.
• Reduced claim frequency and average claim cost due to early warning systems and intervention, ultimately lowering premiums.
• Competitive advantage due to few insurers having made substantial moves in this area, with XYZ seeing this as an opportunity to position themselves in the market.

What's new

Smart home devices monitor and alert homeowners to potential accidents. For example:

• Water leak sensors provide warning of pipe leaks and potentially prevent water damage.
• Sensors detecting smoke and temperature fluctuation provide early warning of fire.
• Motion sensors and security cameras provide early warning of security breaches.

XYZ would require access to sensor data to incorporate this into their underwriting and pricing models. This poses additional requirements for the insurer including:

• Gaining consent to access the data from the policyholders that have sensors installed.
• Incentivising existing policyholders that do not have sensors to install sensors and share data. These could include premium discounts and subsidisation of sensor and installation cost. XYZ may also investigate the expansion of this initiative by partnering directly with sensor manufacturers, in a bid to improve consistency of source data over time.
• Partnering with smart home technology providers to access aggregated sensor data to complement the data sourced from their portfolio. External sources to supplement XYZ's own data are important as only a small proportion of XYZ's customers have sensors installed.

How this influences actuarial considerations

XYZ's actuaries would be part of the team leading this initiative across three key areas:

Strategic role and expanding of cross-functional engagement

Strategic role and expanding of cross-functional engagement: This initiative will require close collaboration between the actuarial team and other parts of the business (particularly data, underwriting, analytics and legal teams).

Some of XYZ's actuaries may take on a larger product ownership role to leverage their perspective across pricing, profitability and regulation. The actuarial team would also work to bridge knowledge gaps in using new machine learning tools and techniques and have been working with experts internally and externally to build up their knowledge.

Initial investigation and updates to traditional actuarial methods and workflows

Sensors provide a new form of data with AI-based techniques used to analyse and interpret that data (similar considerations apply to other sources like satellite imagery). As a result, the following are some key areas to consider for this new technology:

• 
  
  • Enhanced pricing models (with potential flow-on effects to reserving): A new data source such as real-time smart sensor data provides a richer understanding of individual homeowner risk characteristics. More granular pricing and risk differentiation is needed to allow XYZ's pricing team to better tailor premiums.
• 
  
  • Premium discount structure: As part of quantifying the reduction in frequency and severity stemming from the installation of smart sensors, XYZ's actuarial team need to determine a discount structure that fairly translates expected reductions in claims cost to reductions in premium. XYZ is also seeking to incentivise sensor installation by subsidising the cost to policyholders - their actuarial team can investigate what an appropriate level of subsidisation would be and the extent to which this may be offset by a reduction in risk cost.
• 
  
  • Bias/fairness: There is potential for bias against homeowners that do not install sensors to be introduced in XYZ's models. Ways to address this may include introducing discrimination protections against low-income customers or against older homes and homeowners that may not be in a position to install sensors. The evaluation of models for bias, directly and through indirect proxies, is expected to naturally extend to the new data source.

Regulatory compliance and responsible AI

XYZ's strategy includes a commitment to pass savings on to customers where the installation of sensors translates to a lower risk profile, all else equal. A formal structure for monitoring this corporate commitment can be put in place to ensure that savings are passed on to customers. With new sensor data that may include automatic and real-time components directly from homeowners, XYZ's actuarial team is expected to work closely with the rest of the business to ensure appropriate data privacy safeguards are in place, including prevention of leaked sensitive data by AI models.

In summary

The growing adoption of home sensors is expected to result in a lower claims frequency as more claim events are prevented, and a lower claim severity due to early warning that allows action to limit damage to be taken sooner. Both of these elements should ultimately result in lower premiums to homeowners.

Home sensors are still gaining traction, and although acquisition and installation costs have been a barrier to wider deployment by homeowners, insurers like XYZ are in a position to accelerate the uptake through cost subsidisation and premium discounts, along with the necessary protections for customers' data and privacy.

Have you overcome a complex business challenge or pioneered an innovative solution using AI? Share your story to promote how actuaries are using AI.