Resources

Not every wildfire spreads through forests. Some move house to house, block to block, driven by wind, heat, and proximity. As these events become more common, insurers are asking sharper questions about how to assess risk in densely built environments.

Z-FIRE was designed to account for the real-world conditions that enable this type of fire behavior, capturing the structural, spatial, and environmental factors like building density, defensible space, mitigation strategy, climatology, and slope that contribute to urban conflagration.

What Is Urban Conflagration?

Urban conflagration refers to a fire that spreads rapidly through a densely built environment, jumping from structure to structure rather than moving solely through vegetation. Unlike traditional wildfires that primarily burn forests or grasslands, these events are driven by building materials and the adoption of fire mitigation practices, the spacing between structures, and proximity to areas that are known to be challenging for firefighters.

Wind and terrain—particularly steep slopes—often accelerate the spread, pushing fire fronts into residential areas and compounding the risk of widespread destruction. For insurers, urban conflagrations represent a uniquely challenging peril, where fire behavior is shaped as much by the built environment as by natural fuels. Without a model built to capture these dynamics, they’re nearly impossible to assess accurately.

We’ve seen what happens when wind-driven embers reach tightly packed communities. In fires like Marshall, Lahaina, and the recent Palisades and Eaton Fires, wildfires didn’t stop at the wildland-urban interface—they became full-scale urban conflagrations.

These events are stark reminders that wildfire risk doesn’t end at the edge of the forest and that densely populated areas can produce concentrated, catastrophic losses.

For insurers, these events raise a critical question: Where are the next high-concentration loss scenarios hiding in the portfolio?

When urban conflagration strikes, damage rates spike, and PML exposure can increase exponentially. Identifying the neighborhoods where construction patterns, terrain, and fuel conditions create the potential for that kind of fire behavior is no longer optional; it’s essential. 

At ZestyAI, we’ve been asking, and answering, these questions for years. The Z-FIRE model was designed to do exactly that.

Built to Reflect Real-World Urban Risk

From its inception,  Z-FIRE has accounted for the drivers of urban conflagration. Its two-level architecture combines neighborhood-level dynamics with property-specific characteristics, offering a detailed and scalable view of wildfire risk—even in densely built environments.

Level 1 (L1): Neighborhood Risk Score

L1 predicts the likelihood that a property will fall within a wildfire perimeter. It does this by analyzing climatology, historical wildfire behavior, terrain, fuel type, and wildfire suppression ratings to understand where fires are likely to start, spread, and grow.

Two variables are particularly important for identifying urban conflagration risk:

• Fuel Type, which accounts for both vegetative fuels and the built environment. In densely developed areas, clusters of structures can serve as fuel, particularly when combined with slope, dry conditions, or limited defensible space.
• Wildfire Suppression Rating (WSR), which indicates areas where fire suppression is likely to fail due to access, water availability, or firefighting capacity. Rather than relying on WSR alone, Z-FIRE factors in proximity to high and very high WSR zones, enabling it to capture risk spillover into nearby neighborhoods.

What sets Z-FIRE apart is how these variables interact. The model doesn’t assess them in isolation—it evaluates how multiple risk factors compound one another. This layered, interaction-driven approach allows Z-FIRE to surface hidden vulnerabilities that simpler, one-dimensional models often miss and accurately identify regions with high conflagration risk.

• L1 evaluates wildfire risk based on local terrain, fuel types, and building density. Critically, Z-FIRE incorporates two critical variables that are key indicators of conflagration risk. Both “fuel type” and proximity to areas with a High or Very High wildfire suppression rating.
  • Fuel Type accounts for both vegetation and for developed land characteristics. While vegetative fuel and its management are key for wildfire, building density can become a large contributor to conflagration risk if combined with the other high-risk factors. It is important to remember that Z-FIRE allows for unlimited variable interaction; a high-density neighborhood can be at high risk of other factors, contributing to conflagration risk.
  • Wildfire Suppression Rating (WSR) reflects the risk that a high-intensity fire may become impossible to manage by firefighters. Because those areas tend to be located in the WUI, Z-FIRE does not only rely on the WSR score, but also uses the distance proximity to high or very high WSR.

Level 2 (L2): Property-Specific Risk Score

At the structure level, L2 evaluates both nearby vegetation and building fuel density, a key driver of structure-to-structure ignition. This metric, validated by the Insurance Institute for Business & Home Safety (IBHS), helps Z-FIRE model how a single ignition can escalate within tightly packed neighborhoods, even in the absence of natural fuels.

Z-FIRE was trained on real-world events with clear urban conflagration patterns. In the 2017 Tubbs Fire, for example, flames spread deep into Santa Rosa, destroying dense subdivisions like Coffee Park. Similarly, recent fires such as Palisades and Eaton moved rapidly through built-up areas, where tightly spaced structures provided a continuous path of fuel.

Case in Point: Palisades and Eaton, CA

In the recent Palisades and Eaton Fires, Z-FIRE’s predictive accuracy was once again put to the test. Our analysis showed that over 91% of the affected area was already classified by Z-FIRE as high or very high risk based on Level 1 (L1) neighborhood scores.

Notably, none of the impacted areas were categorized as “Very Low Risk,” a strong validation that Z-FIRE captured the inherent vulnerability of these communities well before ignition.

Looking at Level 2 (L2) property-specific scores, the correlation between predicted risk and actual destruction became even clearer.

Structures with the highest Z-FIRE risk scores were 50% more likely to be destroyed compared to those with the lowest scores in the same fire footprint.

These destruction rates align with the model’s fundamental architecture: properties with denser surrounding structures, minimal defensible space, and combustible fuels nearby face a dramatically higher likelihood of loss during a wildfire.

The model identified risk in this community not simply based on topography or vegetation, but because of its underlying urban structure, including building materials, spacing between homes, and neighborhood density. These are the very factors that drive vulnerability to conflagration.

Z-FIRE was designed to capture these dynamics, reinforcing its value as a forward-looking tool for rating, underwriting, and mitigation planning.

A Model Informed by What’s Happening on the Ground

Wildfire conditions are constantly changing, and Z-FIRE stays current by continuously incorporating the latest ground-truth data, including updated fire perimeters, confirmed loss locations, and vegetation conditions.

While the core model remains stable and fully approved for rating and underwriting, this steady stream of fresh data ensures that Z-FIRE reflects the most recent fire seasons and emerging risk patterns.

The 2023 and 2024 fire seasons reinforced what we’ve long understood: fire behavior is increasingly shaped by the built environment. Z-FIRE continues to perform as expected across a wide range of scenarios, including structure-to-structure ignition in densely built areas.

These results reaffirm the model’s ability to provide timely, reliable insights to support carrier decision-making in a rapidly evolving risk landscape.

Available Now, for Those Who Need It Most

For carriers looking to better understand and underwrite wildfire risk, whether in traditional WUI zones or increasingly vulnerable urban neighborhoods, Z-FIRE offers a tested, approved, and field-proven solution.

‍Built on over a decade of confirmed loss data and designed to capture the drivers of urban conflagration, Z-FIRE supports smarter decisions across pricing, underwriting, mitigation, and reinsurance.

Importantly, Z-FIRE is approved for underwriting and rating by Departments of Insurance (DOIs) across all western states, with Oklahoma recently added to the list. Carriers can deploy the model today with confidence, knowing it meets regulatory standards while delivering granular, property-specific insights to support risk selection, pricing, and mitigation strategies.

If you're looking to validate wildfire risk insights on your own book of business, we invite you to put Z-FIRE to the test. Our team can run a targeted evaluation to show how the model performs across your portfolio, highlighting risk segmentation opportunities and identifying properties most vulnerable to structure-to-structure spread.

Get in touch to schedule an evaluation and see how Z-FIRE can strengthen your wildfire strategy today.

ZestyAI today announced the launch of ZORRO Discover™, a generative AI agent that delivers instant insights into regulatory filings and competitive market activity.

Solving a $2.5 Billion Regulatory Research Problem

Regulatory research remains one of the costliest inefficiencies in insurance, with an estimated $2.5 billion spent annually on manual reviews. The process consumes millions of hours across fragmented workflows, as filings often stretch into the thousands of pages, slowing competitive analysis and strategic decision-making.

To solve this, ZestyAI built ZORRO Discover on a proprietary generative AI pipeline, leveraging large language models optimized for insurance-specific content, including filings, objections, responses, and regulations from all 50 states.

Using agentic AI, the platform proactively navigates and synthesizes regulatory data, delivering a fast, intuitive experience tailored to the structure and language of insurance filings.

Accelerating Research by 95%

Early adopters of ZORRO Discover have found that insurers can cut research time by an average of 95%, saving thousands of work hours and millions in costs.  

“For too long, regulatory filings have buried critical intelligence under layers of complexity,” said Kumar Dhuvur, Founder and Chief Product Officer of ZestyAI. “Through years of working closely with regulators, we’ve come to deeply understand these challenges.”

“ZORRO Discover puts that intelligence at insurers’ fingertips so they can act faster, stay compliant, and make smarter decisions in an increasingly competitive market.”

How Teams Use ZORRO Discover

ZORRO Discover can be used by compliance, actuarial, product, and strategy teams to accelerate competitive research and reduce regulatory risk. By replacing manual reviews with a consistent, repeatable process, the platform helps teams improve research accuracy, focus on high-impact work, and stay ahead of regulatory changes.

Key Capabilities

• Plain-language search: Analyze and compare filings across carriers, states, and time periods using intuitive queries—no deep regulatory or technical expertise required.
• Zero in on critical details instantly: Transform complex filings into executive-level insights for faster, more informed decision-making.
• Spot market opportunities early: Track shifts in rates, rules, forms, and guidelines to anticipate industry trends and act proactively.
• Run side-by-side comparisons: Evaluate pricing strategies, regulatory objections, or model usage over time and across competitors. Example: Compare windstorm mitigation discounts across three major carriers in Florida.

Expanding to Additional Lines of Business

ZORRO Discover launches with a focus on homeowners insurance and is rapidly expanding to support regulatory and competitive intelligence across auto, commercial, and additional lines of business.

Request early access to ZORRO Discover here.

The Arizona Department of Insurance and Financial Institutions has approved the use of ZestyAI’s Severe Convective Storm suite, including Z-HAIL™, Z-WIND™, and Z-STORM™, giving insurers access to property-level risk insights.

Rising Storm Losses Across Arizona

In Arizona, the most frequent and damaging thunderstorm losses stem from strong, straight-line winds, especially downbursts. These sudden, localized wind events can exceed 100 mph and often cause significant roof and structural damage, particularly in vulnerable homes.

In 2024 alone, Arizona experienced 67 days of severe storm events, according to NOAA. These included high winds, large hail, tornadoes, and damaging downbursts, highlighting the growing volatility of the state’s monsoon season and the increasing threat to both rural and urban communities.

How ZestyAI’s Models Improve Precision

ZestyAI’s models are designed to address these risks. By analyzing how local climatology interacts with individual property features, the platform predicts both the likelihood and severity of storm-related claims, offering a more accurate alternative to traditional ZIP code or territory-based approaches, which often miss critical risk signals.

Each model is trained and validated on extensive real-world claims data and provides clear, transparent explanations of the key factors driving each score, enabling more confident underwriting and rating decisions.

Key Model Capabilities

• Z-HAIL: Predicts hail damage risk and claim severity using property-specific attributes like roof complexity, historical losses, and accumulated damage, identifying which homes are most likely to file a claim, even within the same neighborhood.
• Z-WIND: Combines AI-generated 3D analysis of roof condition, complexity, and potential failure points with local climatology to deliver pivotal insights into property-specific wind claim vulnerability and severity.
• Z-STORM: Predicts the frequency and severity of storm damage claims, including hail and wind, examining the interaction between climatology and the unique characteristics of every structure and roof.

Regulatory Confidence in AI-Driven Models

Bryan Rehor, Director of Regulatory Affairs at ZestyAI said,

“Arizona is a key market for many of our carrier partners, and one of the most rigorous regulatory environments for model review. This approval is a strong signal of confidence in our technology and our commitment to transparency, precision, and regulatory alignment.”

Arizona marks the 19th state to approve ZestyAI’s Severe Convective Storm suite—and the fourth model approved in Arizona overall, joining Z-FIRE™, ZestyAI’s AI-powered wildfire risk model already in use across the state. With elevated risk across multiple perils, Arizona’s adoption of ZestyAI’s models reflects growing regulatory confidence in AI-driven models that enable smarter underwriting, precise pricing, and stronger resilience in the face of extreme weather.

At our 2025 offsite, we had the rare privilege of hosting Scott Stephenson, former Chairman, President, and CEO of Verisk, for a candid fireside chat with our CEO, Attila Toth.

More than a guest speaker, Scott has been a mentor to our founders and a steady source of insight and encouragement on our journey. 

Over two decades, he led Verisk through extraordinary growth: doubling revenue, quadrupling market cap, and shaping one of the most respected analytics companies in the world. Named one of Forbes’ Most Admired CEOs, Scott helped define the modern insurance ecosystem.

But what stayed with our team long after the session wasn’t the accolades. It was his clarity, humility, and deep conviction that great companies aren’t defined by vision—they’re defined by execution.

Here are three takeaways that stuck with us:

Trust is a survival skill

Scott opened with the story of the 1949 Mann Gulch wildfire, where only 3 of 16 elite smokejumpers made it out alive. The team leader lit an escape fire and lay down inside it. He shouted for the others to follow, but no one did.

“They were a team on paper,” Scott said, “but they’d never trained together. They didn’t trust each other enough to act as one.”

The lesson? In high-stakes environments, technical talent means little without trust. Success depends on how quickly and cohesively a team can move together.

Thinking isn’t doing

Smart teams often confuse planning with progress. As Scott put it:

“Thinking about something, even deeply—and talking about it with others—feels like you’re doing the work. But you’re not.”

Insight without execution is just potential energy. The companies that thrive aren’t the ones with the most ideas; they’re the ones that ship, iterate, and improve with discipline.

Leadership is a mindset

Scott began his career as an engineer. Leadership wasn’t on his radar—until curiosity led him to explore it. His message: leadership isn’t a job title, and it’s not reserved for MBAs. It belongs to those who take responsibility, act with principle, and stay grounded in reality.

Why it mattered

Scott didn’t come to ZestyAI to celebrate the past. He came to challenge us. To remind us that in an industry facing rising losses, regulatory pressure, and rapid change, progress isn’t driven by buzzwords or strategy decks. It’s driven by clarity, trust, and the discipline to do the hard work.

That message hit home. As we scale ZestyAI, our ambition isn’t just to innovate; it’s to lead with purpose. To build the new standard for risk analytics. One that helps insurers price with precision, expand access to coverage, and strengthen the resilience of entire communities.

ZestyAI, the leading provider of AI-driven risk analytics for the property and casualty insurance industry, today announced it has secured a $15 million credit facility from CIBC Innovation Banking.

The facility enhances ZestyAI’s financial flexibility and reinforces its strong balance sheet as the company scales adoption of its climate and property risk models, delivering over 31 million property risk assessments in 2024, more than double the volume from 2023, and on pace to exceed 50 million in 2025.

“We’ve built ZestyAI with long-term discipline and a clear mission: to help insurers navigate changing risks with confidence,” said Attila Toth, Founder and CEO of ZestyAI.

“As demand for trusted, property-level risk insights continues to grow, this capital infusion enables ZestyAI to further invest in supporting this rapid growth and become the industry standard for property risk analytics."

Strengthening ZestyAI’s Position for Rapid Growth

Capital that supports scale, product expansion, and customer demand

“ZestyAI is addressing a critical need in the insurance industry with its AI-powered approach to property and climate risk,” said Sean Thompson, Managing Director, California Market Manager at CIBC Innovation Banking.

“We’re proud to support ZestyAI’s next phase of growth as it continues expanding its platform and deepening its impact across the insurance ecosystem.”

ZestyAI’s footprint has expanded rapidly across the insurance landscape, with growing adoption among admitted carriers, MGAs, the E&S market, FAIR Plans, and public-private partnerships. Over the past year, the company secured 33 new and expanded customer partnerships, including 26 brand-new clients, and launched four new products.

Accelerating Regulatory Momentum Nationwide

This growth has been matched by accelerating regulatory traction. ZestyAI’s AI-driven models for wildfire, severe convective storms, and non-weather water damage have now been approved in over 50 regulatory filings nationwide, with recent approvals in Texas, Colorado, Ohio, Georgia, Connecticut, Michigan, North Carolina, Louisiana, and Oklahoma.

Recently reviewed by the California Department of Insurance (CDI) in the rate application process, Z-FIRE™, ZestyAI’s wildfire model, can continue to be filed for rate segmentation and underwriting without further review under the Pre-Application Required Information Determination (PRID) process, effective January 2, 2025.

Fueling a Mission to Expand Access to Coverage

Built at the intersection of property data, climate science, and AI, ZestyAI’s platform delivers precise, parcel-level risk insights that help insurers strengthen underwriting, price accurately, optimize inspections, and expand coverage, especially in catastrophe-prone regions. In 2024, ZestyAI helped carriers and insurers of last resort extend coverage to over 511,000 properties previously deemed uninsurable, advancing its mission to protect the livelihoods of homeowners, business owners, and their communities.

As Colorado’s HB 1182 introduces new requirements for wildfire risk transparency and mitigation recognition, ZestyAI is helping insurers comply with proven, mitigation-ready solutions.

What HB 1182 Requires from Insurers

Recently signed into law, HB 1182 requires insurers to disclose how wildfire risk models impact rates, account for property- and community-level mitigation efforts, notify policyholders annually of their risk scores and available discounts, and provide a clear appeals process for disputed scores.

The new regulations take effect July 1, 2026, across homeowners and condo policies, including admitted carriers and the FAIR Plan.

Carriers need to move early to ensure their risk models, rating plans, and customer communications meet the law’s requirements. ZestyAI’s explainable AI models are already in use by the Colorado FAIR Plan and leading carriers to assess mitigation and support policyholder communications in Colorado and other wildfire-prone states.

"As regulatory expectations around transparency and customer engagement continue to evolve, HB 1182 sets a clear framework for wildfire risk modeling," said Bryan Rehor, Head of Regulatory Affairs at ZestyAI.

"ZestyAI’s models were built with these principles in mind, offering carriers a proven, low-friction way to meet these requirements while delivering a better experience to policyholders."

How ZestyAI Helps Carriers Meet HB 1182 Requirements

• Built for Transparency: ZestyAI’s explainable AI models allow carriers to clearly communicate how risk scores are generated, what factors are considered, and how mitigation actions influence risk.
• Mitigation-Ready Risk Modeling: Property- and community-level mitigation efforts are integrated into risk assessments, supporting discounts and appropriate pricing.
• Consumer Risk Score Disclosures: ZestyAI enables carriers to generate individual risk scores and mitigation factors for annual policyholder notifications.
• Appeals and Score Adjustments: Carriers can update risk scores in real time based on new property information, offering transparency and responsiveness to policyholders.
• Regulatory Alignment: ZestyAI’s wildfire, hail, wind, and severe storm models have been reviewed and approved for use in Colorado, helping carriers meet standards with confidence.
• Multi-State Scalability: Carriers can streamline compliance across Colorado and other regulatory environments using the same ZestyAI platform.

A Clear Path to Compliance in Colorado—and Beyond

ZestyAI maintains strong relationships with state insurance regulators and actively participates in dialogue around evolving requirements.

At the national level, ZestyAI engages with organizations such as the NAIC and leading industry advocacy groups to stay ahead of broader regulatory trends.

Through in-house Rating and Advisory organizations, ZestyAI files its models directly with state departments of insurance, ensuring each solution is rigorously vetted and aligned with jurisdiction-specific standards before reaching the market.