Blueprinting Proptech: Foxy AI (Series II, Part VI)

Blueprinting Proptech: Foxy AI (Series II, Part VI)

Editor’s Note:

The GEM Diamond is a real estate tech think-tank comprised of 500+ founders, executives, VCs, and practitioners. Our mission is to attract the 1,500 most forward-thinking, and diverse, innovators.

Below is a sample of one of the in-depth startup briefings from our Blueprinting Proptech series that members have full access to for only $139/quarter.

Please note that some links below are GEM Diamond Member exclusive.

Without further ado…

Blueprinting Proptech: Foxy AI

Last Updated July 7th, 2022

Using artificial intelligence and computer vision to convert everyday real estate photos into treasure troves of data, FoxyAI’s suite of AI-powered real estate tools, all available through an API, compute quality and condition, renovation costs, and detect objects and materials.

Founded: 2018
Total funding: $2 million
Last funding: Seed led by Hyperplane (November 2021)
Team size: 15-20
Sector: Visual Property Intelligence

GEM Representation: Vin Vomero (co-founder), Dan Hurwitz (co-founder), Stefan Martinovic (advisor), Jane Hood (advisor), and Ben Rubenstein (investor and board member).

Related Transmissions

Learn more: Website // Crunchbase


the foundation to company comprehension

Product: A library of machine learning and computer vision models built specifically for real estate media:

  • General and Interior: A selection of models that determine quality and condition; as well as scene, object and room detection and labeling. For example, the FoxyAI Quality Score determines the quality of finishes using a continuous six-point scale ranging from Luxury to Basic, allowing for organization, classification, and search. Based on the scoring system from the Uniform Appraisal Dataset.
  • Exterior: Detects and labels photo/media of street signs, house numbers, notices and forms, knock evidence, GPS units, maps, lawn grooming, lockbox, boarded windows and doors, and for sale signs.
  • Damage: Detects general damage and more specifically water stains, mold, mildew, moss, standing water, shingles, roof tarps, soffit and/or fascia, and gutter damage
  • Renovation and Repair (formerly a separate product called remodelit): Analyzes the current features, finishes, and quality of the room, highlighting key elements that will impact renovation or repair costs, such as flooring, appliances, kitchen layout, and countertops. Renovation cost data is available for a variety of common renovation projects and can be linked directly to the quality, condition, and object detected in the photo.

Customers: Wide ranging, from government entities, lending institutions and property preservation, and asset management all the way to insurance and appraisal companies, real estate investors, channel partners, and technology companies. Since automating parts of its business, customers have reported reductions in error rate and man-hours spent while seeing increases in efficiency and quality control.

  • Asset and property managers: Assess a distressed property’s condition and understand the extent and type of damage, and also estimate the repair costs.
  • I-Buyers, Power-Buyers, alternative transaction models: Enhance understanding of value to purchase property (improved accuracy of AVM, knowledge before appraisal complete)
  • Tax Offices: More accurate property assessments leads to higher tax collection figures, plus increased ability to fight appeals.
  • Appraisers: Increase speed and accuracy with automation.
  • Real estate investors: Determine where the next “up and coming” neighborhood will be, and complement and enhance AVMs with FoxyAI models to get a more accurate picture of property value.
  • Government Entities: Automate property quality classification, generate efficiencies, reduce costs, and improve quality assurance.
  • Lenders and underwriters: Process and accurately value properties to reduce underwriting risk while also reducing bias. Lower financial risk by ensuring that AVMs and valuation processes incorporate the most recent information powered by AI that continuously learns and improves.
  • Insurers: Better understand a home’s quality and condition in order to reduce risk exposure.
  • Agents, insurers and lenders: Create and/or enhance n homeowner experiences by using FoxyAI’s models in branded applications.
  • Home Improvement lenders: Determine the cost of renovations, improve underwriting efficiency, and accurately determine loan to value ratios.
  • Retailers and advertisers: Help move inventory by allowing customers, designers, and other users to “visualize the possibilities,” e.g., What would a nursery look like with green modern finishes? What would the kitchen look like with white quartz countertops and high-end Swedish cabinetry?

Traction: Increases of 10X revenue and 6X customers Year over Year (YOY), from May 1 2021 – May 1, 2022.  Customers’ experience (across all segments) on average ~50-60x ROI.

Business model: Providing access to any or all models customers choose, monthly recurring subscriptions are charged at a flat amount (enterprise) for a set number of credits. Credits can be used for API calls to the different models, with each model having a different currency rate in credits. Startups are served via the Foxworx program.


the action or process of supplying individuals and other businesses with products or services

As with most SaaS businesses, the company relies heavily on direct sales to certain customer groups, e.g., government entities, insurance customers, asset managers, and property tech executives. Enterprise sales is the long pole of the business.

It leverages a B2B influencer marketing strategy, partnering with and providing tech to folks who rave about its data models and capabilities. FoxyAI’s technology being integral to other companies’ solutions gives FoxyAI free press and awareness. A webinar with ProxyPics is one example of that (see full webinar here).

The clearly sizable investment in content shows social validation, in the form of detailed case studies. And being referenced in AVM News can only help the awareness fight. Recognition as an industry innovator is part of the playbook as well, for instance taking home a 2022 HW Tech100 spot and a 2021 API World Award for API infrastructure.

Supporting in-field technology, e.g., other folks’ apps through its Foxworx virtual innovation lab—a skunkworks for property—is a smart approach to reaching future innovators.


the entities with whom one is competing; the opposition

A range of startups are leveraging AI to provide visual property intelligence and unlock the hidden value of real estate photos. Among them are:

Oda Studio: Property image enhancement and classification tools powered by artificial intelligence and computer vision. 1 million+ photos processed.

RealStaq: Providing compliance expertise and refined national MLS data services, it has processed more than 1 billion intelligent image tags. Automatic property photo tagging, classification, and enhancement through AI. Detects property condition, visual similarities, duplicates, photo compliance. Processes over 10 million photos daily.

Hosta Labs: Uses AI detection of property photos to create property assessments. Main draw is cost and time savings.

Styledod: Mostly concerned with artificially staging real estate photos, but AI image classification of room types is a part of their repertoire. Have staged 50,000+ photos.

Foyer: Built over 400 million listing images, it provides multi-label classification, object detection, and validation.

Purlin: Uses computer vision, machine learning, and thousands of images to find what home buyers like in order to deliver personalized matches.


the action or process of differentiating one’s product or service

Early mover advantage: We know the drill. The more data these models are given, the better they get in terms of accuracy and capabilities. FoxyAI’s several years of customer discoveries are hard to replicate.

Switching costs: With customers embedding its technology and models into their own respective products and services, once it’s in, it’s not likely to be removed. Especially if another competitor is simply touting a lower price point.

Customer Controls: Due to the ability to white-label FoxyAI technology, customers are able to use them with their own applications and AVM’s as if they were internal capabilities, but without developing the expertise in-house.

Unique models: While there are a plethora of visual property intelligence competitors, many of FoxyAI’s models (e.g, FoxyAI Renovation & Repair, FoxyAI Condition Score, FoxyAI Quality Score, and many of the Damage and Exterior Models as well) do not have directly competing models.

Customer Defined Definitions for Scoring: FoxyAI models are built to accommodate customer-defined definitions of their quality scoring and occupancy. This allows customers to seamlessly and efficiently incorporate their “Quality Class” scoring criteria into a use case using the FoxyAI Quality Score.

Breadth of use: Myriad of use cases where customers have leveraged FoxyAI’s data science team to augment their own. Enables its customers to drive efficiency and accuracy across valuations, risk, underwriting, and revenue collection activities.

ROI/Data driven results: An iBuyer pilot ran 19,105 properties through FoxyAI’s Condition Score Model, totaling to approximately 400,000 photos. This AVM enhancement on average brought each of these 19,105 properties ~$3,000 closer to the sales price, accounting for a ~$57,000,000 improvement in property value. Enterprise sales will continue with transparent, profitable results.


extravagant or intensive publicity or promotion

With the use of AI, machine learning, and computer vision in the real estate industry on the rise, more startups have joined the scene to meet the industry’s needs.

I’m on record with the prior belief that (most of) AI and machine learning is a bunch of bullshit. That was a few years ago, but at a high level, I still believe both are massively overhyped. While there are interesting innovations related to each, they are such buzzwords these days that it’s hard to filter fact from fiction.

However, FoxyAI customers within a varied range of sectors (from government to lending to insurance) have seen clear success in automating aspects of their businesses. Rather than skate by on the awe of AI, the product has legs to be a long-term solution for many companies to increase efficiency and reduce error rates.


a podcast for exceptional founders of exceptional startups that are disrupting the real estate industry

View episode 6:


a means or method of predicting future events


I was always bullish on Remodelit, FoxyAI’s original direct-to-consumer product that helped homeowners estimate renovation costs, find design inspiration, purchase products, and connect with contractors—all using photos of their own home. Property photos and a ZIP code were used to provide customized estimates across various types of remodels (economy, stock, semi-custom, custom, and luxury). Once you dove into a specific project, such as a bathroom renovation, it unlocked what was possible. Further, they had taken what I consider a smart approach: a browser extension on top of Zillow’s inventory.

That product is offline and it’s now operating as a B2B model called FoxyAI Renovation & Repair. Everything in that realm is now only relevant to that one model; it is now B2B and if white labeled would be B2B2C.

I remain enamored with the idea of an entire real estate search experience based on what could be, rather than what currently is. As I mentioned in newsletter #25:

…virtual remodeling and staging is an untapped opportunity to better serve home buyers. The vast majority of people don’t have the “vision” necessary to picture homes differently than their current condition/state. How do you search by nice kitchen, a deeply subjective criteria? You can’t. Personally, I believe there is an entirely new search experience in the “fixer upper” category to be realized by someone forward thinking (with deep technology and design expertise). That customer experience will blend the best of all worlds (curated property search, virtual staging, renovation estimates, and 3D predictive modeling).

A search interface on top of hand-curated “fixer upper inventory.” Perusing homes already “virtually renovated” by top designers. Comparing the total cost over 15 years of a 1920’s fixer with a 1980’s in need of a cosmetic overhaul. Those are among the buyer experiences that have me excited.

Whichever forward-thinking entity chooses to undertake that vision will need deep technology and design expertise, and blend the best of all worlds (curated property search, virtual staging, renovation estimates, and 3D predictive modeling).” While FoxyAI doesn’t appear to be going after this, I would be beyond thrilled to see the right B2B2C product in tandem with a partner like Home Depot, Lowes, or even Zillow. Or, perhaps more likely, HGTV. Every household in America already watches HGTV’s various shows (I’m personally a fan of Chip and Jo). Its audience is a massive distribution advantage that would cost a startup hundreds and hundreds of millions of dollars, and 5-10 years, to capture.

Imagine applying FoxyAI’s technology across all available inventory and enabling buyers to search properties based on what they could be rather than what they actually are. That’s the future of search. Perhaps, Zillow or believe that, too—and make this their next acquisition.

Reminder, this was published as part of our Blueprinting Proptech Biz Intel series.

Interested in a GEM Diamond Membership? Click the button below to apply.