Property and casualty insurers have priced risk using postal addresses for decades. An address resolves to a coordinate, the coordinate maps to a zone, and the zone carries a risk classification. This system is still used and assumed to be sound in much of the industry. Geospatial property intelligence data, combining high-resolution aerial and satellite imagery with AI-driven computer vision analysis, is providing new dimensions to the underwriting process, while reducing its costs.
The geocoding gap and a billion-dollar blind spot
Traditional street-level geocoding resolves an address to a single coordinate, typically the centre of an area, not the structure itself. The positional difference between the two can be a matter of metres, but the financial difference can be measured in billions. Recent research on AI has discovered that over a million properties across the US were underpriced for flood insurance alone as a direct result of inaccurate geocoding. A further 600,000 were overpriced. The total unidentified risk from this one peril: approximately $43 billion.
That covers only flood. It excludes fire, hail, and hurricanes. The compounded exposure across all perils and all portfolios is a liability that most P&C books of business carry without fully quantifying.
Building-based geocoding and the emergence of property DNA
Building-based geocoding resolves this by anchoring risk assessment to the actual structure, not the land parcel boundary. When a property's footprint is derived from machine learning applied to high-resolution imagery, underwriters gain a spatial reference that reflects where a building is located relative to flood zones, wildfire-prone vegetation, and elevation contours.
Factors like roof type, roof condition, construction age, materials, and proximity to known hazard geographies are described as the property's DNA. It is a unique, building-level risk profile that no postal code or parcel centroid can replicate. Using the property DNA also reduces the need for physical inspection and the likelihood that premiums require adjustment after binding.
The underinsurance problem has proven that accuracy is crucial in underwriting. Two in every three homes are underinsured by an average of 22%, and in some cases by 60% or more. Correcting the most severely underestimated living area values alone represents a $2 billion opportunity across P&C portfolios, not from growth but from better data.
Generating the hazard risk score
The core advantage of geospatial intelligence in technical underwriting is its ability to produce a hazard risk score that is property-specific and regularly updatable, rather than assigned at the zone level on a slow revision cycle. Traditional risk zone classifications apply a uniform rating across broad geographies, regardless of whether an individual structure sits on elevated ground, has a recently replaced roof, or is surrounded by cleared defensible space. Geospatial AI scoring draws on real-time and historical inputs, like recent weather event footprints, satellite-observed vegetation changes around a structure, and post-disaster imagery captured after a nearby wildfire or flood.
Ninety per cent of insurance executives identify AI as a top strategic priority, and the value of AI in the global insurance market is projected to reach $80 billion by 2032, up from approximately $10 billion in 2025. The carriers that integrate geospatial property intelligence into underwriting infrastructure now will price risk more accurately than those still relying on zone-level maps.
Underwriting, claims, and fraud
The same data that sharpens underwriting decisions also reshapes what happens when a claim is filed. McKinsey had projected in 2020 that by 2030, as many as half of all routine property claims may be resolved digitally. Before-and-after aerial imagery enables automated damage validation. It compares a property's documented pre-loss condition against post-event imagery to confirm damage extent and location without requiring physical inspection.
Fraud remains one of the most severe financial drains on P&C portfolios. Fraud losses currently exceed $122 billion a year, adding an estimated $400 to $700 to the average American household's premiums annually. A carrier holding current, accurate geospatial imagery of a property can determine whether a claimed roof replacement actually happened, or whether documented damage is consistent with the property's known pre-loss condition. The free-roof scams and staged-damage fraud that lead to heavy losses in markets like Florida depend on carriers not having that information. Geospatial intelligence bridges that information gap.
What implementation actually requires
Building footprint data derived from machine learning and high-resolution imagery must connect to existing underwriting systems, rating engines, and claims workflows to function at scale. Adding geospatial data to one part of the process while the rest of the organisation continues on address-level inputs produces internal inconsistency, and can perpetuate the mispricing that the data was supposed to correct.
The technology itself is capable and scalable. The integration architecture may not be. This is why many insurers are working with specialist operations partners rather than attempting to build these data pipelines entirely from their own resources.
How can Infosys BPM provide underwriting support?
Infosys BPM insurance services support P&C carriers in integrating geospatial intelligence across underwriting, claims, and fraud management operations. It helps build the data infrastructure that translates imagery and AI into pricing accuracy and portfolio resilience.
Frequently asked questions
Traditional geocoding resolves an address to a single coordinate — typically a parcel centroid, not the structure itself. Building-based geocoding anchors risk assessment to the actual building footprint derived from machine learning applied to high-resolution imagery. Research shows over one million US properties were underpriced for flood insurance alone due to inaccurate geocoding, representing approximately $43 billion in unidentified risk from a single peril — before accounting for fire, hail, and hurricane exposure across the full portfolio.
Property DNA is a building-level risk profile capturing roof type, roof condition, construction age, materials, and proximity to hazard geographies — data no postal code or parcel centroid can replicate. It eliminates the zone-level uniform rating that assigns identical risk classifications to structurally distinct properties within the same geography. Industry data shows two in three homes are underinsured by an average of 22%; correcting the most severely underestimated living area values alone represents a $2 billion opportunity across P&C portfolios from accuracy improvement, not growth.
Geospatial hazard risk scores are property-specific and regularly updatable — drawing on real-time inputs including recent weather event footprints, satellite-observed vegetation changes, and post-disaster imagery. Traditional zone classifications apply uniform ratings across broad geographies on slow revision cycles, systematically mispricing properties that sit on elevated ground, have recently upgraded roofing, or have cleared defensible space. Carriers integrating geospatial intelligence now will price risk more accurately than those still relying on zone-level maps as climate event frequency and severity increase.
Before-and-after aerial imagery enables automated damage validation — comparing a property's documented pre-loss condition against post-event imagery to confirm damage extent and location without physical inspection. Carriers holding current, accurate geospatial imagery can determine whether a claimed roof replacement occurred or whether documented damage is consistent with known pre-loss condition. Fraud losses currently exceed $122 billion annually in P&C, adding an estimated $400–$700 to the average US household's premiums — geospatial intelligence directly closes the information gap that staged-damage and free-roof fraud schemes depend on.
Substantial. Building footprint data and AI-derived hazard scores must connect to existing underwriting systems, rating engines, and claims workflows to produce consistent output across the organisation. Adding geospatial data to one part of the process while the rest operates on address-level inputs creates internal mispricing inconsistency — perpetuating the accuracy problem the data was intended to correct. The technology is capable and scalable; the integration architecture frequently is not, which is why many carriers partner with specialist operations partners rather than building data pipelines entirely in-house.
