Finding the perfect site for a development project might seem straightforward in the age of data-driven everything. But for real estate developers, that search has long been anything but easy, says Oliver Alexander, CEO of Prophetic Technologies.
Until recently, even basic information about land availability was fragmented or inaccessible. That's what Prophetic Technologies set out to change with its new artificial intelligence tool, SearchAI Intentions. The platform allows developers to describe what they want to build in plain language—rather than relying on dense zoning terminology—to locate parcels that fit their plans.
"The most frustrating thing is how much data you'd expect is available nationally that doesn't exist yet," Alexander tells GlobeSt.com.
"Think of federal transportation planning. How many parcels within 3 miles of a seaport are at least 10 acres? We can tell you that for the first time ever, but it's [only] been about three weeks since we were able to do that. There are just so many pieces of data to track down and report and [make] public — it's like trying to put out a wildfire with a water bottle."
Prophetic Technologies' ZoneAI data engine extracts zoning regulations for more than 23,000 U.S. municipalities, with 99% accuracy, according to the company. The platform uses large-language-model generative AI to bridge the gap between zoning language and developer intent, identifying suitable parcels for specific project types.
"For a reference point, every single city and every single county, with the exception of Houston and a lot of rural counties in Texas, has specific zoning requirements," Alexander says. These govern where development can occur, as well as minimum lot sizes, structure limits and other factors.
"You're having a good day if you go to a rural town with an 80-page book." Most zoning codes, he adds, typically range from 500 to 1,000 pages.
Prophetic's system aims to streamline that complexity.
"We help them find their next development site, we help them automate the viability process for the development," says Alexander. "What am I allowed to build for zoning? What are the densities? Limitations? What are the environmental issues with the site? Wetland? Toxicity?"
That complexity comes in part from the inconsistent way zoning language is written. "Think of someone's house, a single-family detached housing unit, residential living, all the different ways you could describe that," Alexander says, noting that his team counted 9,378 variations in metropolitan codes.
To manage this variation, the system relies on 30 or 40 master categories and about 120 intermediate keywords. Each result is assigned a confidence score, with 80% indicating high confidence and anything below that warranting a closer look.
"The system does the classifying itself, and the humans come in for a very small subset of results that the system is not able to confidently decide," says Alexander.
"Either a senior staff researcher or the [chief technology officer] will make those calls. It's ultimately a function of who is closest to the project. It takes someone with a very specific skillset and a degree in that field to interpret things effectively. At the end of the day, we show the confidence scores to the users. Sometimes the manuals themselves are ambiguous and don't tell you, and it is a guessing game."
© Arc, All Rights Reserved. Request academic re-use from www.copyright.com. All other uses, submit a request to TMSalesOperations@arc-network.com. For more information visit Asset & Logo Licensing.