The use of algorithmic underwriting is increasing across the insurance industry. With enhanced decision-making and improved risk assessments, an algorithmic approach to underwriting can optimize operations for insurers and enhance the experience for their customers.
In this post, we delve into the evolution and advantages of algorithmic underwriting and share our insights on building and scaling an algorithmic underwriting platform.
The Evolution…
Algorithms have always been a part of the underwriting process, but they have generally been restricted to rating. For example, in determining risk factors for car insurance, algorithms, or mathematical formulas, would be used to set rates based on vehicle make, model, driver age, location, and previous history. Whether simple or complex, algorithms have long been our core rating tool in the insurance sector.
The use of algorithms in other areas of the underwriting process has been limited due to fears of overlapping these factors with rate-making or simply due to the lack of data and analytical capabilities in other parts of the underwriting process to make these decisions. Instead, the insurance industry has typically depended on complex rules engines for decisions on risk acceptance, risk tiers, and report ordering.
With advancements in data access and analytics tools, carriers are now rethinking the use of algorithms, employing them either alone or alongside traditional rules engines, to enhance decision-making throughout the underwriting process.
How It Works…
Algorithmic underwriting employs analytical models to automate decision-making in the underwriting process or to provide insights to assist underwriters. For more homogeneous risks, it can fully or partially automate underwriting.
Key decisions made using algorithmic underwriting include:
- Determining if a submission fits the carrier’s risk appetite.
- Identifying key risk characteristics such as the correct SIC/NAIC code.
- Prioritizing accounts based on desirability and winnability.
- Making risk determinations on portions or the entirety of the risk.
Through this approach, carriers can achieve faster risk acceptance or rejection and reduce underwriting workloads. It also helps provide customers with more personalized risk assessments, real-time risk management, and a seamless experience.
5 Advantages of Algorithmic Underwriting
Algorithmic underwriting significantly benefits the insurance industry across five key areas:
- Process Efficiency: By automating the underwriting process, we are seeing algorithmic underwriting reduce processing times by up to 50%, streamline operations, increase testing speed, and simplify the maintenance of complex decision-making systems. Additionally, the automated processes of algorithmic underwriting can help handle an increase in applications reviewed by up to 25%, enabling insurers to increase premiums without additional operating costs.
- Accuracy: The accuracy of risk assessments can be improved through the analysis of more extensive data sets. These analyses help identify patterns and correlations that might be missed by human underwriters alone. With this augmentation of the underwriter’s insight and judgment, errors in risk assessments can be minimized, and fraud can be more easily detected. We estimate fraud losses may be reduced by up to 30% for some insurance companies.
- Price: Pricing decisions can be more accurate by enhancing risk assessments. Algorithmic underwriting helps tailor premiums to individual risk profiles, enhancing customer satisfaction and competitiveness. Additionally, it supports dynamic pricing, adjusting premiums in real-time based on changing risk factors, which we see improving underwriting profitability by up to 20%.
- Proactive Risk Management: Algorithms can help insurers proactively identify emerging risks and adjust their underwriting and risk management strategies. This can help mitigate potential losses, reduce loss ratios, and improve overall portfolio performance.
- Customer Experience: Algorithmic underwriting allows for instant or near-instant decisions on coverage eligibility, pricing, and personalized offers. With predictive and prescriptive analytics, insurers can make real-time, contextualized offers, making insurance more accessible and relevant to the individual customer’s needs. It also makes insurance more attainable for customers or segments that may have been marginalized by the underwriting methods of the past.
Building an Algorithmic Underwriting Platform at Scale
An algorithmic underwriting platform requires a multi-layered approach that considers future scalability. Advanced features needed when developing an algorithmic underwriting platform include machine learning models, real-time risk assessment, and dynamic pricing models. Challenges to consider as you optimize your data and algorithmic underwriting platform include:
- Data Quality and Availability: Data may be fragmented, incomplete, or outdated.
- Model Interoperability: Complex machine learning algorithms used for underwriting may lack transparency and interoperability, making outcomes difficult to explain.
- Compliance: As regulation of algorithmic models and AI increases, insurers must stay ahead of the guidance and adjust models as needed.
- Fairness and Bias: If not proactively addressed, algorithmic underwriting presents the risk of perpetuating unfair practices and historical biases.
- Data Privacy and Security: Algorithmic underwriting involves collecting, processing, and storing large volumes of personal and sensitive data. Securing customer data is vital for compliance and maintaining customer trust.
Success Stories…
We see examples of success with algorithmic underwriting across the insurance industry. In P&C, for example, Ki Insurance leverages AI and algorithms for instant commercial insurance quotes and automated policy issuance. Hiscox collaborated with Google Cloud to develop an AI model that automates underwriting for specific products. Meanwhile, on the life insurance side, Ethos employs machine learning to assess risk and offer simplified insurance applications.
Conclusion
While algorithmic underwriting is not a novel concept in insurance, it is revolutionary in its enhancement of access to new data sources, improved data quality, and better analytics tools. These enhancements allow underwriters to gain insights from other areas of the value chain and extend their capability beyond archaic models or knockout rules.
Despite their sophistication, insurers must be aware of the potential for bias and a lack of transparency in algorithmic underwriting models. Ethics and compliance, including data privacy, consumer protection, and fair lending laws, will pose challenges for insurers to address from the outset.
