ZestyAI has introduced a new AI-powered model designed to assess non-weather-related fire risk at the individual property level. The model, called Z-SPARK, focuses on identifying factors that influence ignition and fire spread, providing insurers with more detailed insight into potential fire losses.
Non-weather fire incidents remain a significant source of property damage. In 2023, these events accounted for approximately $25 billion in losses. Common causes include grills, appliances, heaters, and electrical faults. Despite the scale of these losses, many insurers continue to rely on neighborhood or territory-level averages and limited historical data to evaluate fire risk. However, risk can vary widely between properties, even within the same area. As a result, insurers may face challenges in accurately pricing policies, leading to adverse selection and unexpected losses.
Z-SPARK addresses this gap by applying modern fire science and property-level data. The model evaluates building materials, maintenance conditions, nearby structures, local fire response capacity, and climate factors. It analyzes how these elements contribute to both ignition risk and fire spread.
In addition, the model uses advanced machine learning trained and validated on millions of real fire incidents and verified insurance claims. It predicts both the likelihood of a fire starting and the potential severity of resulting losses. According to ZestyAI, Z-SPARK delivers 30× greater risk differentiation compared to traditional territory-based models.
With property-level insights, insurers can adjust several aspects of their operations. They can align premiums more closely with actual risk, rather than relying on broad geographic averages. They can also support straight-through processing for lower-risk properties and focus underwriting resources on higher-risk cases that require closer review. Furthermore, insurers may gain the ability to write business in more challenging markets with a clearer understanding of exposure. These insights also support efforts to manage concentration risk across portfolios before losses accumulate.
Z-SPARK builds on ZestyAI’s existing modeling capabilities. The company’s Z-FIRE model is already used to assess wildfire exposure. The new model expands this approach to everyday building fires, which represent a major source of insurance losses.
The broader ZestyAI platform includes models for additional perils such as hail, wind, severe convective storms, and water damage. It also incorporates agentic AI tools designed to support insurance operations. Through its ZORRO Discover platform, insurers can research markets, prepare filings, and act on risk intelligence more efficiently.
Together, these tools provide insurers with a more detailed view of property-level risk and support decision-making across underwriting, pricing, and portfolio management.
Get the latest insurance market updates and discover exclusive program opportunities at ProgramBusiness.com.