The annual GW Business & Policy Forum drives public conversation on issues that carry global impact while spotlighting the university’s strength as a research institution. The 2024 forum, focused on “Imagining the Future with AI,” turned attention on real estate thought leadership at GW School of Business.
During a forum “fireside chat,” Interim Dean Vanessa Perry tapped her expertise on racial bias and home ownership while speaking with Bright MLS President and CEO Brian Donnellan, M.B.A. ’02, about AI’s potential impact on the real estate industry.
Perry is known for her high-profile research on the homeownership race gap—only 43 percent of Black households live in a home they own—and how it impacts wealth accumulation. White households in the United States have a median net worth at least 10 times that of Black households, according to a 2020 report Perry co-authored for the National Association of Real Estate Brokers. Donnellan, meanwhile, leads a multi-listing service used by real estate professionals in six states and the District of Columbia. It is reportedly home to real estate’s most popular property-tour website.
Donnellan and Perry spent much of their discussion looking at whether AI could address one of the biggest challenges for the sector: the lack of reliable data. Public records inform the field, meaning the information has to be loaded into real estate databases by hand. “If you go further away from a city, you probably need six months to get that info and get it in a format that is consistent,” Donnellan said.
Perry said that means the data is “old by the time it gets collected… so all the analysis is old. It’s not telling us much about the future.” Donnellan agreed that the dearth of timely reports on real estate sales hinders efforts to design national housing policy.
Can AI help solve some of the data disjointedness? Donnellan said a lack of uniform standards for reporting information is also challenge. Individual real estate professionals do not use the same terms when describing property amenities, for example. Also, in noting that historical data on home values is embedded with racial bias, he said AI only has the potential to correct appraisal flaws if it can first detect bias.
Even more, he said real estate is a “very, very fractured” industry, largely self-regulated and driven more by emotion than by data, making it difficult to predict how artificial intelligence could reshape it.
“It’s the most unusual industry. We represent big cities. We represent mountains and farms,” Donnellan said. “These professionals are different from anyone. Bright MLS is used by 100,000 independent contractors acting in 100,000 different ways.”
Donnellan said AI is integral to real estate searches but noted that it is just one piece of a broader landscape that also includes inspections, repairs, financing and massive amounts of paperwork. He also acknowledged that many real estate agents are afraid they will lose their jobs to AI, just as travel agents did—nearly overnight—once people had the ability to compare and book their own airline tickets online.
In discussing the future role of real estate agents, Perry suggested that real estate professionals may find themselves taking a larger financial advisory role. Donnellan agreed, adding that AI could help agents more effectively use the leads that come in and better serve first-time homebuyers. However, he did not expect dramatic AI-related changes in the sector for at least seven to 10 years.
Perry, like many GW faculty members, brings a practitioner’s eye to her research. Before joining academia, she worked for mortgage giant Freddie Mac, the U.S. Department of Housing and Urban Development (HUD) and the U.S. Consumer Financial Protection Bureau.
The availability of reliable real estate data and predictions about AI’s impact on the sector were also discussed during a forum panel that examined emerging AI policy issues in housing and financial services and a closing keynote address by Jeremy Wacksman, chief operating officer of the tech real estate marketplace Zillow.
The discussion on housing and financial services was moderated by John Carlson, vice president of cybersecurity regulation and resilience at the American Bankers Association. Panelists were Stefano Pasquali of BlackRock and Nicholas Schmidt of SolasAI. Pasquali uses AI for mathematical modeling in the investment arena. At SolasAI, Schmidt specializes in algorithmic fairness. SolasAI provides companies with software that tests fairness.
The panel participants pointed out that AI has been used in financial modeling for housing for some time.
“AI makes [financial modeling] cheaper and allows them to use more data and allows them to use larger models,” Pasquali said. “Ten years ago, it wasn’t scalable. Now it is.” He also said the user experience has improved with AI; users can interact with platforms more efficiently.
Schmidt, meanwhile, noted that computer power has improved and there is more data and analytics to support decision making. It is possible to add additional complexities to make the model even higher performing, but this is where governance needs to be addressed, he said.
“This stuff has been around for a long time but now it’s really gathering steam,” Schmidt said. “I see these algorithms being used in pretty much every part of financial services, starting with marketing then pricing and underwriting of loans. And now machine learning could go to service, which is the area that feels the most dangerous.”
Schmidt recommended that companies roll out AI use in a limited way, perhaps for product-use enhancements such as building out marketing content. He said companies also need to be wary of third-party vendors who may build models that may not include what is promised.
“Don’t adopt these things quickly, adopt them smartly,” Schmidt cautioned.
Carlson discussed how regulatory requirements for financial services and, to a lesser degree, housing could also have unintended consequences for AI. Schmidt countered that these guardrails may help manage the risk presented by AI. He added there may be no need to employ complicated mathematical models to achieve desired results.
“The practices and regulations that the financial industry have are actually very good. One question is whether people need to use such complicated models,” Schmidt said. “Maybe they can use just machine-learning models. That sort of work is really valuable.”
Echoing concerns raised in the discussion with Perry and Donnellson, Carlson asked the panelists what can be done to mitigate data drift or lack of data.
“Data is more important than ever,” Pasquali said. “The problem of data quality is that the quality cannot be ignored.” He also said the challenge with data migration is that it is sometimes done by third parties.
Schmidt said new data streams can provide information about customers who were closed out of credit markets, such as people without FICO scores who may be creditworthy. “But the question is whether that’s valuable or not,” he said. “You may be bringing more people into the system … but missing data doesn’t usually help people.” He said the more important thing is to understand how racial discrimination enters the model.
“Are minorities sufficiently represented in the data?” he asked, noting inaccuracies in current data. “The error rate for identifying the gender of women of color is 35 times more than that of white men.
“Data based just on minorities will incorporate all the inequities. Understanding this data is essential,” he added.
The panelists said speed alone should not propel AI usage, and they agreed that fairness, accountability and transparency are not at odds with innovation. They also maintained that legislation affecting AI needs to advance a principle-based approached and offer strong guidance on how regulation can be enacted.”
Pasquali called on academia to help develop governance tools.
In closing the forum, Zillow’s Wacksman discussed how AI is reshaping the behavior and power of consumers within the real estate market. In 2006, the company launched “Zestimate,” a machine-learning model to help people understand the value of their homes and put the company at the forefront of innovation. Wacksman believes AI could add additional streamlining to the complicated home-buying process.
“It takes months to buy a house, thousands of documents and it costs tens of thousands of dollars,” Wacksman said. “People describe buying a house as more stressful than planning a wedding or getting a job.” He said Zillow wants to bring all the disparate parts of a house sale transaction together.
Wacksman said Zillow looks at AI with three goals in mind: to give customers what they want, to embrace business opportunities and to innovate responsibly. In particular, he said innovation could bring more accountability to a sector that has a long history of inequality and racial inequities.
According to Wacksman, Zillow has spent years personalizing the home-buying experience. The company started by enabling customers and realtors to understand the value of a home, and now it is working “to personalize the finance experience,” Wacksman explained. Zillow is also experimenting with technology, including virtual reality walk-throughs of properties for sale, a process that is currently too expensive to launch mainstream.
The Zillow executive said generative AI has the power to transform the real estate industry, helping both customers and real estate agents. The challenge will be ensuring that everyone is comfortable using the same technology.
“It’s daunting and exciting to see the pace of new change. It is measured in days and weeks, not months and years,” he said. “We think it could be as significant as the Industrial Revolution.”