Inman

Mike DelPrete: Zillow and Redfin’s AI tools aren’t that big a deal — yet

Ted Irvine | Inman

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.

When it comes to ChatGPT in real estate, it seems everyone is trying to either understand it, implement it or emulate it.

The OpenAI language model, used by more than 100 million people worldwide, has turned heads due to its convincing conversational prowess, its knack for processing vast amounts of information (although not always accurately), and even its ability to perform specific tasks like identifying issues in human-written code. It’s also spawned competitors like Google’s Bard chatbot, which seem destined to expand this technology’s impact even further.

If applied to real estate’s largest databases — such as those maintained by search portals like Zillow and Redfin — generative AI has tantalizing potential, Mike DelPrete believes. So why isn’t the real estate tech strategist all that impressed by these companies’ first steps into this arena?

Mike DelPrete joined Intel by video chat Thursday to discuss this new frontier for real estate tech firms. His thoughts below were edited for length and clarity.

Intel: Zillow and Redfin are starting to roll out plug-ins testing what these new generative AI features could be used for. Is this a big deal for real estate search? Or is there more untapped potential here that they haven’t yet explored?

DelPrete: No. What we’ve seen so far is not a big deal at all. At best, it’s a proof of concept that, in my opinion, has pretty marginal utility for consumers. But it’s a first step in this process. So it’s a necessary, unexciting building block to fulfilling the potential of AI.

It’s like building a great new mansion and pouring a concrete foundation. It’s not exciting, right? But it’s necessary to get there.

The plug-ins that Zillow and Redfin have put out there basically reproduce what you can already do on the websites; it’s just conversational, and like 10 times slower. There’s really limited consumer utility around that.

And I think it’s more about them being able to say they released a ChatGPT plugin before their earnings call, which is exactly when these things came out. So unfortunately, this is what we have to face with public companies and earnings and wanting to sway investment sentiment, is stuff like this. 

What do you think the exciting applications might look like once they get to the next level?

Generally, putting generative AI on top of multiple proprietary datasets — that’s the most exciting thing.

An example: “Show me all the three-bedroom houses for sale where I can walk my kids to school,” or, “Show me all the homes for sale in this neighborhood that are at least 10 percent below the median home value and haven’t been previously listed in the past 6 years.”

That’s really specific stuff, and it would take a human being a long time to come up with, if at all. And they’d have to cross-reference multiple datasets. But AI on top of those datasets can figure that stuff out really freaking quick.

Zillow’s also doing some stuff around machine learning and image recognition, so yeah. You can get pretty specific.

The ultimate irony is that with no inventory and nothing for sale, it doesn’t really matter. We’re talking about fine-tuning these very specific houses, but what people are really asking right now is, “Show me anything for sale. For the love of God, just show me something for sale.”

For a business like Zillow, you’ve described these generative AI chatbots as more of a “top of the funnel” play. Can you elaborate on that? How does generative AI actually play into the Zillow business model?

Zillow got its start by getting all the listings online — so it democratized the real estate search process — and then also the Zestimate. The Zestimate was a fun tool. It’s like a toy that consumers could play around with that’s related to real estate, and it didn’t exist before that.

Right now, generative AI and ChatGPT, it’s kind of the same thing: It’s this fun tool. It’s fun to play around with.

If a consumer’s doing a house search, I think by and large, most consumers are going to get the same result whether they’re using ChatGPT and AI, or just searching Zillow. And that’s because it’s real estate. People are willing to put in the time and energy and effort required to look at all the inventory for sale. And consumers are willing to deal with a higher degree of pain to get that inventory.

If anything, it’s going to be a fun tool and a fun toy for consumers to play around with to augment the home search process. So that’s why I say it’s at the top of the funnel. Like the Zestimate, it’s another way for consumers to interact with the real estate market in a fun and interesting way.

And if you attract consumers with a tool like Zillow did with the Zestimate, then you have those eyeballs at the start of the process, and you can guide people through the journey and monetize those eyeballs, which is exactly what Zillow’s done for the past 15 to 20 years.

With a question like, after it’s given you an answer, “Are you ready to talk to a real estate agent,” for instance?

Yeah, totally. I think that this idea that AI — these AI chatbots — are going to replace agents seems far-fetched.

Zillow didn’t replace agents. Zillow augmented agents. More people are using agents now than ever before. AI can’t tour a home with you.

So I think it’s more about a fun tool at the top of the funnel for consumers, and then after that, perhaps a productivity-enhancer for agents. Again, if you’re sitting on top of a lot of datasets and there’s a lot of busy-work, perhaps AI can do that.

And what we’ve seen right now is automating writing property listings. Doing some work on the purchase agreement or the legal documents. Or maybe you can throw AI on top of a CRM database and provide some unique insights or some help in staying in touch with the database. So I think that’s where the productivity enhancements for agents really come into play.

As far as who the winners and losers might look like in the marketplace here, let’s say some tech-savvy startup gets in on AI home search and does it better than Zillow. They’re years ahead of Zillow on understanding the tech and making the right consumer features. How do you see a scenario like that ultimately playing out?

I don’t see it playing out.

Zillow’s not the be-all, end-all, most fantastic user experience. Many people talk about Redfin as having a superior technological experience. I’ve looked at and studied and researched — and worked out — real estate portals around the world. They’re not the best user experience. Other startups have better user experiences and more intuitive ways to search.

But what a lot of people forget is it’s really the inventory that matters. That’s it. People are willing to deal with pain to do this. So I think the idea that some scrappy startup is going to get access to the same level of data that Zillow has, current and historical, throw an AI chat search experience on top of that that’s materially better than Zillow — not just different, but better, faster, more efficient in a real way that affects consumers — is incredibly unlikely.

Zillow wins. That is kind of reductionist. Zillow is not guaranteed to win here, but it’s very hard for Zillow to lose in this space because of the eyeballs they already have and because of the engineering horsepower they can throw at this.

Email Daniel Houston