New in Proptech | Fall 2026: What Shipped and What It Means

Proptech companies are starting to sell AI as the better option instead of just the cheaper one. In July, Thesis Driven’s Brad Hargreaves described the Great Proptech Convergence: as AI makes software easier to build, vendors keep expanding into one another’s territory. At first, companies competing to become the all-in-one solution for entire departments got there by studying manual workflows and replacing them with AI. The pitch was primarily about headcount: automate enough work and a firm can cut its analysts. Recently, though, we’ve started seeing more proptech companies going beyond the idea of replacing humans with AI, and moving to deploy the technology to do things humans weren't able to do before.

This is a quarterly roundup of what’s new and most interesting in proptech, drawn from hundreds of conversations I’ve had with proptech CEOs and CTOs as Thesis Driven’s technology liaison. Seven companies made the cut. Five of them launched major products in the past three months (Geolava, Zoneomics, Giraffe, Pirros and Acres); the other two (Andes and HABU) are early-stage companies testing new business models. 

Each entry covers what the company does, what’s new about it, why it matters, a quote from its leadership, and ends with the bear case, since most of these bets are early.  

Geolava

What it does: Geolava builds AI models of the built environment from satellite imagery, street-level photos and spatial data to forecast physical change.

What's new: On October 26, Geolava launched Terra, what it calls the first world model built for real estate. It’s easy to confuse with computer vision, but the two do different jobs: computer vision describes what a property looks like today, while Terra, Geolava says, learns how properties evolve over time, simulates possible scenarios, and predicts how a building could change under different conditions. 

Pointed at a road, a computer vision model reports "no pothole," or flags a hazard if a crack has opened. Terra studies how similar roads have deteriorated, then simulates traffic, weather and maintenance scenarios to predict where and when this one will fail, and which repairs would prevent it.

Why it matters: Forecasts like that turn property data from a report into an underwriting input. In its first commercial uses, Terra has predicted major physical changes with 86.5% accuracy, measured against changes later confirmed on the ground, according to Hantz Févry, Geolava’s cofounder and CEO. Asset managers and investors already use the model to anticipate neighborhood changes, underwrite acquisitions, monitor portfolios and project CapEx.

Geolava also recently signed a contract with the Florida Department of Transportation to model how natural disasters will affect its infrastructure and to schedule predictive maintenance of roads and sidewalks.

Founder quote: "Most of the attention in physical AI goes to robots. But before you can act on the physical world, you need a model of how it changes. Infrastructure decays slowly and expensively, and that's exactly where seeing the future is worth the most" – Hantz Févry 

Bear case: The technology is early, and like other AI models, a world model of the built environment will never reach 100% accuracy. At 86.5%, about one prediction in seven is wrong, which matters far more when underwriting an acquisition than when scheduling road repairs. Terra also depends on enormous volumes of data that is expensive to acquire, thin outside dense metros and often restricted by licensing terms.