The machines are home: Will AI transform residential real estate tech?

The AI revolution has proptech boosters proclaiming that in the future, real estate’s truism will go from “location, location, location” to “data, data, data.”

The acceleration of AI technology and hype around a new generation of chatbots, like OpenAI’s ChatGPT and Google’s Bard, suggest potential to reshape residential real estate technology. Proponents see better machine learning tools creating rapid and more accurate real-time valuations for homes, generative AI spitting out marketing material, data mining identifying owners likely to sell, and savvy chatbots changing broker-agent and agent-client relationships. Those in the business of selling homes have already begun tapping existing AI tools to speed up rote tasks like finishing property listings and paperwork.

But despite the gold rush mentality infecting proptech – more than $8 billion in equity and debt investments were made in the first half of 2022, according to Houlihan Lokey – realizing AI’s potential will hinge on the strategy of established players, investor sentiment, and whether or not major residential firms will assimilate the tech into their workflow.

“Now is when it gets tactical,” said Seth Siegler, who oversees technology for virtual brokerage eXp Realty, which has nearly 75,000 agents across the U.S. “Some companies will jump ahead, some will fall behind. But everybody needs to be cognizant of what’s just happened.”

The general mood among proptech VCs is that there isn’t yet an explosion in AI-related funding, but that it’s becoming a hot commodity. Startups are promoting any AI-related feature, no matter how insignificant, and increased investment in the space is expected later in the year.

Dan Wenhold, a partner at real estate-focused venture capital firm Fifth Wall, sees customer service and chatbot programs, such as Elise.AI, already in use by AvalonBay and Stonehenge, as the true opportunity. Seed and pre-seed investment in nascent AI proptech firms will take off in the later half of the year, Wenhold predicts, with this year’s total investment “exponentially higher” than 2022.

Incentivized to spot and profit off the latest trend, venture capital firms lean toward techno-optimism. Both Fifth Wall and Metaprop were big cheerleaders of the use of crypto/web3 tech in real estate, a sector where hype peaked in 2021 and quickly dissipated.

There seems to be more caution around AI.

“We are not getting overly excited about AI, much in the same way that we didn’t get overly excited about blockchain and tokenization startups,” said Jeffrey Berman, a principal at real estate VC firm Camber Creek, which just closed a new $100 million proptech fund. “The rush to throw money at any startup that says ‘we’re using generative AI technology’, we arch an eyebrow.”

Others, though, are sticking their neck out. Agya Ventures co-founder Kunal Lunawat, who sees transformative possibilities in the space and says his firm will make investments, sees a real possibility of disruption. Big tech is busy firefighting and coping with layoffs, and the Innovator’s Dilemma — being consumed by the effort needed to keep existing business models alive, only to then be usurped by up-and-coming technology — is a real issue, especially in the listings space. Just as the merger of ChatGPT and Microsoft’s Bing has threatened Google, Lunawat believes a new AI-powered startup can disrupt early real estate tech successes like Trulia, Redfin, or Zillow.

Plunk, a Redmond, Washington-based startup focused on using machine learning and AI to more accurately value homes, aspires to be the “Bloomberg for residential real estate,” according to CEO and co-founder Brian Lent. Plunk’s analysis factors in “at least twice as many” data sources as Zillow’s Zestimate and removes bias, Lent claims, creating a more robust, real-time, and actionable home valuation; a homeowner can upload pictures of their kitchen remodel and get a near-instantaneous recalculation of the home’s market value. Plunk can even serve as a tool for institutional investors seeking out single-family rental properties, Lent said, with The National Association of Realtors and California Association of Realtors both on board as investors.

Agents “bring a knife to a gunfight,” said Lent, who previously worked at Amazon and whose collaboration with Sergey Brin at Stanford helped birth the Google Web Crawler. “They’re competing with Zillow, and people saying ‘Zestimate says this,’ and it’s not really a statement of value. They need more weaponry, more tools to be smarter, to both value the home and be able to educate the homeowner.”

It’s worth noting that Opendoor also touted its superior home-valuation algorithm, but got caught out in the housing downturn: It lost nearly $1 billion in the third quarter of 2022.

Meet the new boss. Same as the old boss

Even some singing the praise of AI’s revolutionary potential will note that incumbents already have the advantage in incorporating the technology, namely huge data sets and customer relationships, the former being incredibly important to training AI models – the bot needs thousands of at-bats to get smart.

Milestones.AI, an Austin-based startup building a home management system, believes it can feed data-heavy profiles of residential property into generative AI to suggest when to sell, the right time to call a repairman to fix the furnace, and even connect agents with great leads. But this concept requires not just great programming, but laborious data entry.

“It’s going to take us 10 years to get to the place where we’re really smart,” said CEO and co-founder Dustin Gray. Part of the solution is working with title companies that close millions of transactions a year to start building the database. So far, the firm has amassed 200,000 users and raised $15.4 million in venture capital.

Without the right dataset, novel AI chatbots can often deliver incorrect answers that can damage the reputation of startups, said Dan Teran, managing partner at New York-based venture firm Gutter Capital.

”I think it is more likely that AI will be applied by incumbents,” added Teran, who said the early hype reminds him of crypto. “Adding AI to a startup pitch is like adding 50 percent to the price tag,” he said.

Zillow, for now, isn’t sounding the alarm about an AI startup’s ability to take its business. Its Zestimate – love it, or loathe it, it’s a thing – has used machine learning since its inception, and launched an AI search feature in January. It plans to continue to refine its listings to incorporate more natural language interactions and ways to do more realistic 3D-tours.

“We’ve had a head start for quite a while and some of our coolest AI stuff has been launched in the last couple years,” said Zillow spokesperson Matt Kreamer.

The brokerage eXp has been experimenting with OpenAI since receiving a beta invite in late 2021, according to founder Glenn Sanford, who came back as CEO at the start of the year. He believes the ability to train ChatGPT, vastly improved and simplified from previous iterations, means it’s practically turnkey, and any firm could incorporate it into a useful AI-enabled bot.

“It’s hard to know just yet if it will be a revolutionary or evolutionary member of the brokerage tech stack, but on the surface, it looks like it will make a great companion for inside sales teams,” said Sanford. “Its ability to quickly answer nuanced questions or proactively ask its own questions will improve consumer experience and lead conversion.”

As more startups compete for funding and firms get pressure to incorporate new tech, the engineering talent gap will prove to be a big challenge: There’s a reason Plunk’s Lent refers to AI and machine learning talent with real estate experience as “pink unicorns.” But perhaps the biggest question will be how much this tech lives up to the hype – a question not even ChatGPT can answer.

“The final frontier that we’re trying to cross is whether people are actually making meaningful investment-based decisions and moving millions of dollars due to this technology,” said Fifth Wall’s Wenhold. “Right now, it’s all about figuring out what is smoke and mirrors or vaporware, versus what is an actual product.”