Inman

With Fannie Mae’s adoption, AI’s role in real estate is solidified

Craig C. Rowe; Canva

This report is available exclusively to subscribers of Inman Intel, a data and research arm of Inman offering deep insights and market intelligence on the business of residential real estate and proptech. Subscribe today.

Artificial intelligence is not far from becoming an entrenched component of the real estate transaction, as evidenced by Fannie Mae’s formal announcement of its use in property appraisals in its June 2023 Appraiser Update.

The GSE (Government Sponsored Enterprise) stated in the paper that computer vision, a form of AI that can read and extract actionable data from still images, can be used to correct appraisals that didn’t sufficiently detect differences between subject properties and their comparables.

While Fannie Mae requires appraisers to include interior and exterior photos of properties they’re appraising, it only requires one exterior photo for each comparable property. Without interior photos for comps, it can be difficult to detect when condition and quality ratings in appraisal reports don’t match up.

To fill in the missing information, Fannie Mae acquired interior photos of comps from other sources such as Multiple Listing Service (MLS) listings, and used image recognition technology to analyze the condition and quality ratings in appraisal reports.

Fannie Mae said that it analyzed a million appraisals using computer vision and found that “the test identified a small subset with a high probability of erroneous C [condition] or Q [quality] ratings.”

“Appraisal experts in our Loan Quality Center reviewed those reports and found the model prediction was 98 percent accurate. This enabled us to identify appraisal defects much more efficiently than our other processes, including many defects that previously were impossible for our technology to triage.”

As a result, Fannie Mae said in its report that it “plans to deploy this technology in CU [collateral underwriting] and in our standard selection process for post-acquisition loan quality reviews in the near future.”

Restb.ai, the real estate industry’s go-to provider of computer vision technology, now a part of a large number of proptechs and multiple listing services, is seemingly well-positioned for this role. The company already provides mortgage originators and appraisal management companies (AMCs) a GSE-compliant image validation solution with its computer vision technology, according to a Nov. 29 press release.

Tony Pistilli, general manager of valuations for Restb.ai, said in the release that the GSE’s attention to computer vision will foster substantial interest from lenders and appraisers.

“Image recognition is now a must-have,” he said. “If Fannie Mae is doing it, appraisal providers and lenders need to be doing it too.”

The company released a product specifically for valuations in August of this year, and shortly after partnered with another company in the appraisal technology space, Bradford Technologies. Restb.ai’s computer vision and machine learning technology will be integrated into Bradford Technologies’ report quality control processes, Inman reported.

Restb.ai was way ahead of the recent industry push into AI integration, recognizing more than five years ago that real estate was the most applicable use of its technology.

The Barcelona-based technology company considered how it may overlap with other lines of business, but the value that imagery holds in residential real estate was too obvious to ignore, especially given the sheer volume of it. In essence, without computer vision, untold terabytes of data are left out of the transaction ecosystem. With it, still images become much more valuable than their use in marketing.

Nathan Brannen, chief product officer for Restb.ai, emphasized that fact in a July 2023 interview with Inman Intel.

“There are about a million photos uploaded every day in the U.S., just in MLSs,” he said. “Someone may buy a home without seeing it, but no one buys a home without seeing a photo of it. They contain so much information.”

Restb.ai is not the only player in this space. New Zealand’s ListAssist is working with California’s TheMLS to assist its members with automating property input, creating listing descriptions and ensuring image compliance. RealScout was early to use the technology to improve home search, allowing users to search by and compare interior features, a growing use among more prescient proptechs and tech-savvy agents and brokerages.

The adoption and proven capability of image recognition within the GSE space indicates its time has come. It’s no longer fringe or bleeding edge; it’s on the cusp of widespread acceptance, and, in turn, the real estate transaction may finally become more efficient as a result.

“Lenders and AMCs can immediately benefit by reducing the number of appraisal corrections, mitigating loan repurchase risk, and improving appraisal turn times and increasing appraisal quality by adopting our computer vision solution,” said Brannen in the release. “Computer vision can be a trusted source for quality and condition ratings, identifying property damage, home features not mentioned in the appraisal, and most importantly, protect the lender and appraisal provider by validating information in the appraisal.”

Email Craig Rowe