AI to take on a larger role in homebuying

Artificial intelligence will play a bigger role in helping consumers buy and sell homes, as it moves from behind the scenes to front and center in a whole range of areas, according to members of a panel at the annual conference of the National Association of Real Estate Editors in Las Vegas.

Zillow, the country’s largest real estate portal, began offering “natural language” searches on its mobile app to bring up more precise results. Rather than setting filters across a limited set of criteria, like the number of bedrooms and ZIP Code, consumers can speak or type what they want.

For example, a consumer could search for a three-bedroom house with a brick exterior built in the 1950s or 1960s with a large backyard near a public park in Denver.

“Artificial intelligence will have a tremendous benefit to the real estate industry,” said Jasjeet Thind, senior vice president of AI and Analytics at Zillow, on Wednesday.

Zillow has been deploying machine learning and artificial intelligence since 2006 in its Zestimates, which are automated estimates of a home’s value, Jasjeet said. Initial efforts generated a pricing error range of 14%, but that is now down to 2%.

The use of AI has spread into a host of other applications. The pandemic made 3D showing popular, allowing consumers to do virtual walk-throughs of a home online. Zillow takes it a step beyond, using AI to create floor plans and square footage estimates for each room based on photos.

AI will also allow real estate sites to engage with buyers and sellers of homes before they are ready to talk to a broker, answering their questions and boosting their comfort level, said Rob Barber, CEO of ATTOM, which maintains large data sets on real estate markets.

He also sees it being increasingly used to help consumers see what a given room might look like with new furniture or different appliances.

ATTOM uses artificial intelligence to generate risk estimates for insurers and lenders, which in turn can help generate more customized insurance premiums, Barber said.

The panelists said it is unlikely AI will eliminate the need for human brokers or the commissions they charge, which can run in the 5% to 6% range for a seller.

But they could reduce the friction or sticking points in real estate transactions, which is a good thing, said William Holmes, head of agent partnerships at Opendoor, a provider of instant offers.

AI, while making great strides, is far from foolproof.

“AI is like an idiot savant. It doesn’t know the truth,” said Glenn Phillips, CEO with Lake Homes Realty, on a different panel on Tuesday.

The models remain highly suggestible and will carry forward the biases of the coders. Bad data results in bad answers. And AI remains prone to hallucinations, or fabricated answers, that sound plausible but have no basis in reality.

And even when AI is working as it should, natural language searches still depend heavily on how well queries are written.

“You won’t be replaced by AI,” Phillips said. “You will be replaced by people who are better at using AI than you are.”