Zumper's CEO explains why the rental platform built a ChatGPT plugin instead of embedding AI into its own app. The plugin lets renters search with natural language like "I want a southwest-facing apartment near Whole Foods" and get listings. The team shipped it in three weeks, betting that meeting users on ChatGPT is essential as search surfaces shift.
So you did mention AI, right? And we couldn't go to the chat without talking about that. And I know I shared the post as far as like your, you know, your plugin. So it looked really cool. Could you kind of, you know, share a little bit more about how you guys are experimenting with AI?
Yeah, we kind of thought when, when the APIs started opening up, and I mean, we've used AI and machine learning for years at Zumper. And so some of this stuff for many marketplace founders isn't new. Um, so when, when kind of the, the, especially the OpenAI kind of APIs became available to us, uh, we kind of, to oversimplify, thought about it 2 ways. Do we either leverage this on platform, so on Zumper, to create a better user experience, or do we go and farm this and go fish in a different pool to say, well, ChatGPT or Bard or Anthropic are going to become new surface areas where people will start a rental search? Do we need to go meet them there? So the first one was, do we make Zumper's product better? The second one was, do we go and fish for new users to bring them to Zumper somewhere else? And for various reasons, we picked the latter. We said, we believe that ChatGPT and Bard and a few other kind of cases are going to be places where people will start their rental search. Even if they end up on Zumper, they will start there. And it's existentially important for us to go there. So what that basically baked into was we built a plugin for ChatGPT-4. It's really cool. It doesn't replace like a map list interface that's immersive like we have or Airbnb has in their segment. But it's hard to believe that like ChatGPT-4 Bard is not going to cannibalize search. And Google obviously know that, hence Bard. And I think as a marketplace founder, it's existentially important to go and meet users where they are. And never assume that you have full distribution unless you're picking those next channels, of which many will fail, but 2 or 3 will not. And it reminds me a little bit of building for iOS and Android early. Like Zumper built some of the first ever like Android and iPhone apps. We even built for like the iPhone and Google Wear early on, even though we knew that the watch probably wasn't the most natural like marketplace extension. But you just have to take your swings. And I think we'll find with AI chatbots, That the plugins that Zumper built will maybe one day just become almost like an XML feed that we send these chatbots and the plugins are just a stepping stone to just sending them a sitemap. That's kind of how I can see it evolving. And consequently, as a large consumer marketplace, you just have to work with these chatbots and actually they're a phenomenal user experience for renters who have very specific queries like, hey, I want to live in Southwest San Diego, west-facing windows near a Whole Foods. Well, you can find that on Zumper through our ChatGPT-4 plugin if you look for it. And I think this is just like the beginning.
No, certainly. And it's definitely cool to see. But, you know, I noticed when you were sharing it on Twitter, you know, you got pretty detailed in it, right? So, you know, is this something that you're spending a lot of time, you know, with yourself, or could you share a little bit more about, you know, within kind of your team or, you know, the org?
For transparency, so we've probably got like 50 to 60 people on our product design and engineering team. So we're still pretty lean given how big our audience is. And we built our plugin with a team of like 3 or 4 people over 3 weeks. So what's amazing about the plugin is that you leverage the power of OpenAI or the future plugins that we'll launch. And it's quick. What's not amazing about the plugin is you leverage the power of OpenAI and the future ones we'll build, and it's quick, meaning it, I don't want to say it's commoditized, but our competitors can build this quite quickly too. And so there are fast follows, I think. So yes, I spent a lot of time with it. I was just, as the CEO, I get an unfair advantage in like access to the right people at some of these companies. And so my job's to go, think forward, unlock the doors, and then send my team who are far smarter than I am in terms of building these products into them. But yeah, I was involved with my team in the product development and testing and launching it. And my general kind of takeaway as a marketplace founder at the growth stage is when I first, you know, heard about AI, it was existentially terrifying. It was just like, shit, we're gonna lose the front ends and, you know, everyone's gonna go to these things. I think I've reframed the way I think about it. If I'm Zumper, or I'm Kayak, or I'm Thumbtack, you know what, actually, what's valuable about your company is still your supply. It's the piping of those listings and the way we order them and the way we refresh our feeds. And every second of every day, our rental feeds are doing very sophisticated things underneath the waterline that no one can see. If there was a startup tomorrow that launched a rental startup on AI, I think it would scare me less because those feeds took us a decade to integrate and scale. And ultimately, if you have more inventory, you win. That was a lesson at the very beginning of Zumper, and it's a lesson now. And I think given these LLMs are now open to everyone, I think the marketplaces at scale, for better or worse, have an enormous advantage still because they can use exactly the same LLM. On top of an inventory database that's 10 times larger than anyone else's. And so you could argue this is just going to entrench existing marketplaces if you took that opinion. Time will tell.
