Zestyai Inc v. Aon Re Inc

1. Case Identification

• Case #: IPR2025-01360
• Patent #: 11,030,491
• Filed: July 30, 2025
• Petitioner(s): Zesty.ai, Inc.
• Patent Owner(s): Aon Re, Inc.
• Challenged Claims: 1-20

2. Patent Overview

• Title: Automated Property Feature Assessment Using Machine Learning
• Brief Description: The ’491 patent discloses systems and methods for automatically analyzing imagery of a property to assess its features and conditions. The technology uses machine learning algorithms to extract property characteristics (e.g., roof type, siding type) and classify their condition (e.g., new, damaged) from aerial or terrestrial images, primarily for insurance risk evaluation purposes.

3. Grounds for Unpatentability

Ground 1: Obviousness over Gross - Claims 1, 6-11, and 16-20 are obvious over Gross.

• Prior Art Relied Upon: Gross (Application # 2015/0186953).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner argued that Gross teaches a comprehensive system for assessing the condition of property features for insurance purposes that renders the independent and several dependent claims obvious. Gross’s system trains a “Classifier Engine” using neural networks to analyze property images, identify structural elements (e.g., roofs, façades), determine their characteristics (e.g., shingle type, number of stories), and classify their condition (e.g., new, damaged, missing). Petitioner asserted this maps directly to the core limitations of independent claims 1 and 11, including accessing images, applying boundary information via a "building envelope," extracting features, and applying machine learning. Furthermore, Petitioner contended that Gross’s disclosure of using "multiple classifiers" for better accuracy and customizing them for different purposes would have made it obvious to a person of ordinary skill in the art (POSA) to use a "first set" of algorithms for characteristics and "another" algorithm for conditions, as recited in the claims.
  • Motivation to Combine (N/A - single reference): Petitioner asserted a POSA would have understood that the image processing techniques Gross described for training its classifier were inherently intended for use in the subsequent assessment process itself. Thus, the claimed invention was presented as an obvious implementation of Gross's own integrated teachings.
  • Expectation of Success: A POSA would have had a high expectation of success because the challenge involved applying Gross's disclosed training and analysis techniques to its disclosed assessment framework, which is the intended and logical purpose of the reference.

Ground 2: Obviousness over Gross and Furukawa - Claims 2-5 are obvious over Gross in view of Furukawa.

• Prior Art Relied Upon: Gross (Application # 2015/0186953) and Furukawa (Patent 6,970,593).
• Core Argument for this Ground:
  • Prior Art Mapping: This ground addressed dependent claims requiring the boundary information to be a "shape map" (claim 2) that is used for selecting images that most closely match the boundary (claims 3-4) and for cropping images (claim 5). Petitioner argued that while Gross teaches using boundary information ("building envelope"), Furukawa explicitly discloses using map data containing the "external shape of buildings" to match with and correct satellite imagery. This map data, which plots boundaries and shapes of structures, is functionally identical to the claimed "shape map." Furukawa further teaches comparing image boundaries to map boundaries and, if they match, using that information to proceed. This satisfies the limitations of selecting images based on a boundary match. Furukawa also discloses cropping a three-dimensional image based on this boundary data by setting the area within the external shape as the candidate area for analysis.
  • Motivation to Combine: Petitioner asserted that Gross identified a clear need in its system: the requirement to confirm tentative property addresses that are tagged to images to ensure accuracy. Furukawa provided a well-known solution to this exact problem by using map data with building shapes to verify locations in images. A POSA would combine Furukawa's shape map verification method with Gross's property assessment system to solve Gross's stated problem, thereby improving the accuracy and reliability of the system.
  • Expectation of Success: The combination involved integrating a known data verification technique (Furukawa's shape maps) into a system that explicitly required such verification (Gross), which would have been a predictable implementation for a POSA.

Ground 3: Obviousness over Gross and Davis - Claims 12-15 are obvious over Gross in view of Davis.

• Prior Art Relied Upon: Gross (Application # 2015/0186953) and Davis (Patent 10,755,357).
• Core Argument for this Ground:
  • Prior Art Mapping: This ground addressed method claims requiring the application of a "risk profile" to determine "damage risk" (claim 12), including specific risks like fire (claim 13), modeling property vulnerability (claim 14), and considering neighboring structures (claim 15). Petitioner argued that Gross teaches the foundation for this by disclosing the development of "correlations" between property condition ratings and insurance metrics (e.g., claim severity), which constitutes a "risk profile." To the extent Gross is not explicit, Davis teaches a risk analysis system for insurance that uses stored rules—an explicit "risk profile"—to estimate the likelihood of damage based on property features. Davis provides examples such as rules specifying that dead trees increase fire risk, which maps to claim 13. Davis also teaches estimating a "likelihood that damage will occur," which maps to the "modeling vulnerability" limitation of claim 14. Finally, Gross teaches assessing "appurtenant elements" like fences and garages, which satisfies the "neighboring structures" limitation of claim 15.
  • Motivation to Combine: A POSA seeking to enhance Gross's general insurance risk assessment system would have been motivated to incorporate the specific, rule-based "risk profile" methodologies taught by Davis. Combining Davis's explicit risk analysis rules with Gross's automated condition assessment framework would create a more efficient, accurate, and granular system for insurers to quantify property damage risk, which is the shared goal of both references.
  • Expectation of Success: Both systems utilize image analysis to assess insurance risk. Therefore, integrating Davis's specific risk determination rules into Gross's broader condition analysis platform would have been a straightforward and predictable implementation for a POSA with a reasonable expectation of success.

4. Relief Requested

• Petitioner requests the institution of an inter partes review and cancellation of claims 1-20 of Patent 11,030,491 as unpatentable under 35 U.S.C. §103.