Insurers are losing the climate fight - can AI and data turn it around?
Extreme weather is hitting harder and more often. California wildfires and severe convective storms made the first half of 2025 the second costliest on record, per Swiss Re and Munich Re. Risk models anchored in historical averages can't keep pace with live, shifting hazards.
The fix isn't a tweak. It's a move to real-time underwriting. AI, geospatial analytics, and dynamic climate data are quickly becoming table stakes for how carriers score risk, price, and allocate capital.
Legacy models are breaking - precision underwriting is the new standard
"Extreme weather events are happening with greater severity and frequency," said Megan Kuczynski, founder and CEO of ClimateTech Connect. "Technology… is really critical to precision underwriting."
Look at Swiss Re's CatNet integrating UK-based Fathom's advanced flood modeling. That's the pattern: plug specialized, science-grade models into carrier workflows. As Kuczynski noted, "They're former hydrologists and they really know their stuff." Domain expertise is being baked into tooling, not just reports.
Risk appetite is being rewritten by tech - and transparency
Reinsurers are stepping back from high-risk markets. In response, capital allocation is going data-first. One example: the Connecticut Insurance Department's partnership with First Street Foundation, giving every homeowner property-level climate risk data. That level of visibility pushes market behavior, not just consumer awareness.
Regulatory programs matter too. Alabama's roofing grants help homeowners harden structures, which changes long-tail loss potential. Expect reinsurer appetite to follow the money where mitigation is measurable and monitored.
First Street Foundation is a useful benchmark for property-level flood, fire, wind, and heat insights.
Mind the climate data gap - NOAA is still the backbone
NOAA remains the baseline for many climate models. Pullbacks or delays on federal datasets pose real risk to model development and calibration. "There is a private sector initiative to take over… or partner with NOAA to make sure that those data sets are preserved," said Kuczynski.
If your ingestion pipeline relies on NOAA, build redundancy now. Mirror critical datasets, validate refresh cadences, and set vendor SLAs for continuity.
NOAA NCEI is the starting point for many hazard baselines.
Chronic and systemic risks are now underwriting inputs
Sea level rise, prolonged heat, grid and infrastructure stress - these slow burns are forcing updates to exposure views. "Everything that you cited, just the rising temperatures, are the reason for these growing weather events," Kuczynski said.
Health impacts on outdoor workers are entering the discussion, surfaced at a recent Connecticut Insurance Department climate event. Expect more products, exclusions, and pricing tied to heat exposure, labor safety, and business interruption.
Nature-based mitigation belongs in models
Coral reef restoration that breaks wave energy. Prescribed burns that reduce fuel loads. These aren't just feel-good projects - they alter expected loss and should be reflected in rating plans and portfolio accumulations.
Measure the mitigation effect, then tie it to pricing credits, eligibility, or capacity. If it doesn't move loss cost, it's philanthropy. If it does, it's underwriting.
From reactive to preventive insurance
Insurance has been built to respond. That's changing. "Not one industry can solve in a silo," said Daniel Kanuski of Marsh McLennan, framing the push for tech, regulators, carriers, and consumers to coordinate.
Kuczynski called it "an ecosystem of co-beneficiaries." As carriers pilot predictive adaptation and prevention tech, the overlap between underwriting, risk engineering, and public policy gets tighter.
What to implement in the next 6-12 months
- Stand up a real-time hazard pipeline: Ingest live wildfire, flood, wind, and heat feeds; bind them to property IDs; refresh daily for quote/bind and in-force monitoring.
- Adopt specialized peril models: Evaluate partners like CatNet + Fathom for flood; validate performance against your claims history before scaling.
- Property-level transparency: Provide applicants a climate scorecard at quote. Where risk is high, require mitigation steps for eligibility or pricing.
- Mitigation-linked pricing: Tie roof age, fortified standards, defensible space, elevation, and flood barriers to explicit credits and underwriting rules.
- Dynamic portfolio steering: Update aggregates and reinsurance cessions quarterly (or faster) based on new hazard data and mitigation uptake.
- Chronic risk factors: Add heat stress days, sea level projections, and infrastructure fragility into long-term pricing and reserving assumptions.
- NOAA contingency: Mirror critical datasets, set vendor backup feeds, and add data QA checks to detect stale inputs.
- Public-private alignment: Prioritize markets with grants (e.g., roofing programs) and community mitigation - capacity follows measurable loss reduction.
Vendors and data to evaluate
- Flood: Swiss Re CatNet with Fathom-integrated modeling for high-resolution flood views.
- Property-level risk for underwriting: Fora for assessments that inform eligibility, pricing, and required mitigation steps.
- Consumer-facing risk transparency: First Street Foundation tools for homeowner education and expectation-setting.
- Geospatial stacks: Platforms that combine satellite, lidar, parcel data, and building attributes with fast APIs.
Team and skills - build the bench that can ship
- Blend talent: Hydrologists, fire scientists, and climate modelers paired with data engineers, actuaries, and product underwriters.
- Embed squads: Stand up peril-focused squads (flood, fire, wind, heat) with clear loss and quote-to-bind KPIs.
- Upskill underwriting: Train teams on geospatial data, model assumptions, and how to explain mitigation-linked pricing to agents and customers.
If you're building AI capabilities internally, consider structured upskilling paths for insurance roles. A practical starting point is these job-specific AI course tracks.
The takeaway
Historical averages won't save loss ratios in a live hazard environment. Real-time data, science-backed models, and enforceable mitigation are the levers that matter.
Move fast on ingestion, validation, and pricing changes. And partner where it accelerates outcomes - because the market is already rewarding carriers that can see risk sooner and act before the storm hits.
Your membership also unlocks:
