Case Study

What if insurers could make better and faster decisions about policies, underwriting and pricing?

CLIENT

Global Insurer

Industry

Insurance

LOCATION

Europe

AI-Based Risk Modeling and Pricing for a Global Insurer

A global insurer sought to make more informed and timely business decisions surrounding policy conditions, underwriting practices, and pricing strategies. They struggled to access data and analytics, and it was tedious when they could.

We developed a real-time risk analytics and pricing solution with predictive AI modeling, allowing the insurer to make timely and data-driven decisions. The solution uses machine learning to model risk and expected losses and provides risk-based pricing with dynamic assumption settings. The platform provides real-time reporting and tracking of exposure, risk, expected losses, and overdue debtors.

Modeling and Pricing

The insurer’s modelling and pricing tool is a user-facing platform with a dashboard interface. It’s used by various teams to quantify, manage, and report on risk. Primarily, the tool is used by two groups: underwriters to help make data-driven decisions when underwriting and pricing risks; and management to access real-time reporting and intelligence.

The underlying data is pulled directly from the data warehouse, and relevant custom calculations and models display the desired outputs. The platform features:

  • Real-time tracking and reporting of exposure, risk, expected losses, and overdue debtors
  • Ability to slice data with drill-down functionality to an individual debtor level
  • Risk and expected losses are modelled using machine learning
  • It generates risk-based pricing with dynamic assumption setting
  • Ability to analyze the effect of policy adjustments on losses and pricing
  • Impact of different future economic conditions on risk and losses

Solution Architecture

  • Data is pulled from internal (policy admin system, finance system) and external (macroeconomic data, rating agency data) sources into a data warehouse.
  • Machine learning prediction models (credit risk model, economic model, loss model, pricing model) use input data from and write predictions back to the data warehouse. These models are deployed on Azure cloud and run daily.
  • A cloud-based visualization tool connects to the data warehouse and displays model results and other insights from the data.
  • Data refreshes daily

“The speed at which they understood our problems and created tangible solutions was impressive. The tool was effectively built over three months […]. The use was immediate and required only limited fine-tuning thereafter.”

Head of Underwriting

Let's talk about your next big project.