MW How artificial intelligence could change the way you buy or sell your next house
By Aarthi Swaminathan
Will AI help people navigate an increasingly difficult housing market?
With temperatures running into the 80s in her suburban neighborhood in Dayton, Ohio, Katie Hill decided during the height of the pandemic that she wanted a house with a swimming pool. Hill's two kids - going stir-crazy during the government-mandated shutdown to stop the spread of the coronavirus - needed a respite, and public pools were off-limits.
Hill, 45, owned her four-bedroom house, which had appreciated in value between 2012, when she bought it, and 2020, and had a mortgage rate of around 4%. She decided to take action by walking up to her neighbor across the street, a man close to retirement, to ask if he had an interest in selling his home - which had a pool.
At this point, the housing market was going bananas. Bidding wars, soaring home prices and mass investor purchases all compounded into a pandemic-era buying frenzy. Hill's neighbor said he was keen, and was already thinking about selling - but wanted to wait until he retired, if she was willing to wait.
"That experience got me thinking that people should be having more conversations about their property plans," Hill told MarketWatch in an interview. "If we understand what people's plans are, then that's where there's an opportunity."
The encounter planted a seed in Hill's brain that then evolved into her setting up a new company, which she called Unlisted - a software platform that is driven by artificial intelligence to connect owners of off-market properties with people interested in buying them.
To be sure, her strategy of knocking on her neighbor's door to ask if he was selling was something real-estate agents and investors have done since time immemorial. But Hill believed that artificial intelligence and machine learning could offer significant value to the process of buying and selling homes, particularly homes that are off the market.
Hill is far from the only person thinking about how AI could impact real estate. From trying to predict when and where people are likely to sell their homes, to what kind of outreach would be most effective, real-estate investors - who buy and sell houses relatively frequently - have already begun to use technology to make smarter bets.
But a far more important application of this technology could come when people start using AI for transactions involving homes in which they reside. That technology could potentially upend the traditional method of listing and selling homes and even the business model of brokers, whose business is already under assault due to changes to traditionally lucrative commission rules. At a time when the American housing market is grappling with historically low inventory, the ability to accurately predict when a homeowner will sell is prized intelligence. With the housing market feeling increasingly dysfunctional, using AI to make better residential real-estate decisions could end up being one of the Best New Ideas in Money.
"In the past couple of years, we've seen 300% growth just with our AI product," said Greg Clement, CEO of Realeflow, an AI-driven software provider for people in the residential real-estate market.
There are different types of AI, but in the real-estate industry, the ability of AI to use existing data to predict patterns over time and provide forecasts - as well as generative AI's ability to create images, text, video and more based on user prompts - has unleashed a realm of possibilities to shake up an industry that has generally been a slow adopter to technology.
Aside from data points such as where a house is located and the kind of income its homeowners make, Realeflow also includes variables such as what kind of magazines the homeowners subscribe to, the age of the children in the house and even the type of car they drive, as a measure to predict the likelihood of them selling their home. "We knew that certain things trigger people wanting to sell property," Clement said.
There can be no doubt that there is plenty of AI hype, and the issues facing the American residential real-estate market appear to be structural and enduring. "AI goes in waves, and we're kind of at a teenage phase of generative AI where we're like, 'This is really cool ... but none of us are using it in every facet of daily life,'" Nicholas Stevens, vice president of product, artificial intelligence at real-estate company Zillow Group $(ZG)$, told MarketWatch.
But companies such as Zillow are starting to use AI in different ways, like detecting users' behavior to help them search and identify more relevant home listings.
The problem
The housing market today faces a unique problem: With a so-called lock-in effect, homeowners with ultra-low mortgage rates are holding out on listing their homes. Consequently, inventory is low, as seen in the chart above.
The number of homes available for sale is measured by a metric referred to as months' supply by the real-estate industry. Unsold inventory in March was at a 3.2-month supply, according to data from the National Association of Realtors, which reveals how few home listings currently exist. A more balanced market is considered to be one where there's 5 to 7 months of supply.
The lack of home listings is pressuring home prices upwards. In March, the median price of a resale home was $393,500, while the median sales price of a newly built home was $430,700.
Home prices haven't significantly dropped in the last year, despite mortgage rates going up as high as 8% and settling around 7% for the time being, which has eroded what buyers can afford. Home prices are also rising faster than wages, the NAR noted.
"Home buyers are frustrated," Lawrence Yun, chief economist at the NAR, said on a call with reporters in March. Even if buyers can afford to buy a home, they face bidding wars given the stubbornly low inventory levels, he added.
With this backdrop, the technology that comes with artificial intelligence offers an edge for both home buyers and sellers.
With the ability to perform routine tasks - from applying for a mortgage to digesting millions of home listings to find something attuned to one's interests - the technology could offer not only convenience, but also speed and savings, potentially.
Feeding home buyers information based on preference
The platforms that some companies currently offer provide a glimpse of what could be possible.
Consider Realeflow: The startup offers a comprehensive approach for a fee of up to $400. Investors on the platform can identify which neighborhoods and which homes out 130 million in the U.S. have the highest likelihood of being sold, and also recommends the best times to send mailers or targeted social-media advertisements to those groups.
At another company, New Western - a real-estate marketplace that lists properties geared at fix-and-flip real-estate investors - artificial intelligence plays a role in curating listings for the website's 200,000 users. "In any given city, there's not more than 10 or 12 properties in a day on that marketplace, because they sell in a day," Kurt Carlton, co-founder and president of New Western, told MarketWatch.
As each local investor has a unique set of interests, the firm uses artificial intelligence and machine learning to monitor their behavior on the platform, and gives them a notification, at speed, when something that fits their criteria becomes available, Carlton explained.
"That's very important because these properties are generally available on our platform for [approximately] two hours," he added. "[Investors] can really focus on three or four properties a month, instead of having to spend an hour going through four or five properties."
That level of precision is slowly becoming more accessible to the public.
Zillow - one of the most well-known real-estate companies in America, and famous for its publicly available home listings and its proprietary Zestimate home-value estimator - is also embracing artificial intelligence.
For instance, when searching for homes on Zillow, instead of plugging in one's budget, how many bedrooms they would like and checking off a series of boxes about their preferences on the website, users can now type in what they want in a more natural language that the company's artificial-intelligence models can respond to, allowing them to find listings more suited to their taste, Stevens, Zillow's AI chief, said.
In other words, the new technology can be used to feed users content and ideas based on their preferences - whether they're looking for ranch houses, two-car garages or a house's proximity to coffee shops.
"With generative AI, we're able to extract information about the images - so if you love that big kitchen, [or] a big backyard, you don't have to tell us anymore," Stevens added. "We can see that based on your interactions and then use that in search ranking behind the scenes."
Current homeowners aren't left out. AI is also improving how home values are being measured via the Zestimate, Stevens noted. Instead of simply basing valuations on static data such as square footage, location or the number of bedrooms or bathrooms, so-called unstructured data enters the fray, based on what the listing images look like and how the listing agent has described them.
'The models don't get far without great data'
But getting to a point where the real-estate industry has at last begun to embrace AI was not easy. Experts who spoke with MarketWatch stressed the difficulty in processing highly disconnected data sets together, a necessity when trying to use artificial intelligence.
"The data is more nuanced in real estate," Zillow's Stevens said. "The models don't get far without great data." Location data, mortgage information, property taxes, flood risk, neighborhood desirability, projected rental income, comparable values - the list goes on.
(MORE TO FOLLOW) Dow Jones Newswires
May 22, 2024 09:08 ET (13:08 GMT)
MW How artificial intelligence could change the -2-
When it comes to adopting new technology such as AI, "real estate has always been a bit of a laggard," Alex Wolkomir, a partner at consulting firm McKinsey & Co., told MarketWatch.
Until now, the industry has seen little need for it. After all, home sales rocketed in recent years due to record-low mortgage rates, and the booming status quo didn't incentivize interest in trying new technologies. On top of that, "a lot of technology that is relevant to real estate was not mature," Wolkomir said, "because so much of real estate is still physical."
Staircase, a New York City-based startup, recently launched its flagship product called ChatMTG. The text-based chat platform offers a 10-minute mortgage-application process that uses simple data and is vetted with AI. Plug in all of the necessary information - from pay stubs to bank statements - and the bot will put together an application.
The goal is to use the bots to eliminate the middlemen involved in the mortgage process, Soofi Safavi, co-founder and chief technology officer at Staircase, told MarketWatch. Those intermediaries "collect a significant amount of compensation for it, [when one can do] that with bots," he said.
Plus, the technology helps Staircase offer a mortgage rate that's 1.25% lower than the standard, Safavi noted - which can total hundreds of thousands of dollars in savings over a 30-year mortgage.
The drawbacks
At a time when the industry is facing headwinds due to mounting problems in the commercial real-estate sector, slower rent growth, rising home prices and high interest rates, offering to save folks some money seems like an attractive proposition.
But as with every promising and emerging technology, there comes unintended consequences.
One of those consequences is the amplification of discrimination. Since algorithms get their data from current sources, there is a possibility that they consume significant amounts of flawed or questionable data points, and spew out a less-than-ideal outcome.
In 2016, ProPublica published a deeply reported story about how AI used to predict the likelihood of a person committing a future crime contained racial biases.
In the real-estate sector, algos can also perpetuate discrimination against people of color, as they ingests property data going back decades - including periods when Black people were actively discriminated against and denied property.
"People of color are not represented in the data ... [and so] structural, systemic issues that are well known in housing will be perpetuated by AI systems," Michael Akinwumi, chief responsible AI officer at the National Fair Housing Alliance, told MarketWatch.
Federal regulators are aware of this so-called digital redlining. The Consumer Financial Protection Bureau said it's prioritizing this issue, to make sure that people are protected from algorithmic bias.
That's in addition to automation bias, Akinwumi noted. "Based on your skin tone, some of the algorithms that are behind the facial-recognition technology are over-criminalizing residents of color in certain areas," he said. In 2023, the Washington Post published an investigation that revealed how surveillance cameras were being used to punish and evict residents of public-housing projects.
"As humans, we often overrely on technology and assume anything that is coming out of the system is perfect, is objective. But there's a lack of oversight," Akinwumi added.
And so as AI becomes part of people's daily lives, it also has the potential to "turbocharge fraud and automate discrimination," Lina Khan, the chair of the Federal Trade Commission, observed last April.
'It's become sort of a daisy chain'
But for now, the real-estate industry is still trying to figure out how to actually use the new technology.
Back in Dayton, Katie Hill is trying to figure it out. On her platform Unlisted, Hill brings homeowners and home buyers together - all completely off the grid, or multiple listing services. No home sold appears on public-facing websites like Zillow.
Instead, when a home buyer searches for a property, the company curates, scores and ranks a list of homes specific to what they're looking for, from its database of 121 million homes.
Once Unlisted figures out which homes fit the buyer's preferences, its software figures out how likely it is that those homes will sell - a similar model to Realeflow's.
"The higher the score for that search, the higher the home is ranked and presented to the buyer as a promising opportunity for that specific buyer," Hill explained.
Hill's company also invites homeowners who are identified as having desirable homes that other people would want to buy to sign up for the platform. They don't have to sell immediately; rather, the platform asks when they might be ready to sell. This information is relayed back to the buyer, and feeds into an algorithm that informs how likely other similar homes will sell. "As the responses increase, the algorithms are further informed and refined, and the predictions become more accurate," Hill said.
For homeowners, the goal is for the platform to "signal, subtly, that they're opening [up] to selling," which will eventually help the company match them with eager buyers, Hill said. "It's become sort of a daisy chain where you can start to put together" people who are looking to sell in a couple of years with people who want to buy during that specific time frame, she noted.
So far, Hill says Unlisted has reached out to some 1,000 homeowners in 14 cities - and since October has facilitated the sale of six homes, all off the market.
-Aarthi Swaminathan
This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.
(END) Dow Jones Newswires
May 22, 2024 09:08 ET (13:08 GMT)
Copyright (c) 2024 Dow Jones & Company, Inc.
Comments