Play with it. But basically, it's got ChatGPT built right into the browser. Now, that may not mean a lot, may not impact you that much, but what it is is a gateway to eventually being able to ask your computer to do stuff on your behalf, which could be things like, 'Hey, I just met John Doe at the coffee shop, and I'd like to start a drip campaign for them.' And it will then start to put it into your CRM and start to do things on your behalf. So you're going to basically be able to skip a whole bunch of the learning steps that right now, if you think about where we were two or three months ago, you had to prompt a lot more. Now, the AIs are getting better and better where one prompt, and you can start to actually get real work done. This is in all the models: ChatGPT, Claude, those types of things. Yesterday here at EXPcon, I did a session on AI and building an app, and somebody threw out an idea: they wanted to be able to take an inspection report, upload it, and find out what the costs are to repair a property based on the inspection. I was like, 'Well, could you ask me to build something easier?' Anyway, 30 minutes later, eight prompts in Lovable, we had a platform built that would read an inspection report from any place in the country, break it down to its component parts, write down the prices of each, and give a total cost of what it would take to do the inspection repairs to bring the property back up to where the inspector would have been happy.
And here I was thinking that I was real cute just taking a report and dropping it into a mainline chat and saying, 'Help me make an email to the other agent about the repairs or the credits that need to be done.' This right here just changes the whole game, the whole dynamic, because it can be optimized for different markets and regions and areas. So if I'm sitting in Syracuse, New York, and you're sitting in Washington state, the cost for some of these services might be different. Generally, the cost of lumber is the cost of lumber, but the cost of labor may be different, some materials may be harder to source. That could fundamentally change how any of these agents are having informed conversations, especially in a negotiation on behalf of their client.
Right? Yeah. Because they're going through and getting bids on all this stuff. It takes a while. AI has a pretty good understanding of a lot of things, so it can come up with what it takes to fix the driveway of some concrete cracks. It's seen a lot of data, it kind of knows. So it can say, 'It's going to generally cost around this to fix that, especially in your area.'
We're catching up with the founder of EXP Realty, Mr. Glenn Sanford, here on KGCI Real Estate on Air. We are in the final day from the Miami Beach Convention Center for EXPcon. Where are we in the development of AI? I hear from a lot of agents, especially as I've been milling around, they seem disheartened about AI, discouraged. They feel like they haven't gotten started yet and must have missed the boat. Where are we in AI's development? Related to content streaming, are we still in the Napster era or even before Napster? Are we at dial-up?
The interesting thing about AI is that if you go back to ChatGPT in 2023 or 2024, and you look at 3.5, that was kind of like, 'Oh, this is novel. It'll compose me some tweets, write me a little essay, give me a poem.' Then we went to 4, and it started to become a bit of a thought partner. Now we're at GPT-5, and it's a bit of a beast—it can do a lot. But each time the model is improved, it shortens the time for the next new person to figure it out. The ability for somebody to say, 'I need an inspection app to read an inspection and give me the cost'—if I had tried to do that a year and a half ago, it would have taken a ton of prompts and probably wouldn't have worked. Now I can just do it. Within a fairly short period of time, say the next six months, you'll be able to say to an AI, 'Build me a real estate CRM attached to a real estate website for my community that captures leads, and then for grins and giggles, I'm going to ask you, AI, what else should we build?' The AI is going to say, 'Well, you should add an events calendar, and you should do this because I think it'll increase engagement. Add a social component tied into your social media.' And you'll go, 'That sounds good.' It'll do it because it'll be agentic—it will connect to your social media and start to operate as you. People might be disheartened because they're not super techy, which is that bleeding edge. But within a few months, six months or so, you're not going to need any tech.
So for the agents that maybe tried to hop on ChatGPT back in the 3.5 era and had a hard time with it, what I hear you saying is let not their heart be troubled because it's gotten simpler.
It's gotten more intelligent. I say that we actually put the intelligence in AI. It's artificial intelligence. We can sort of figure out what's good information, what's good feedback from a brainstorming partner perspective. But it learns from more and more people saying, 'That's good, that's bad.' So it figures out what's a better user outcome each time somebody prompts it. Even though you may not have gotten a great response six months or a year ago, and you might have said, 'This is frustrating, AI is dumb as a doornob,' AI is still dumb as a doornob. It surprises me when I spar with it—I'm like, 'Oh, great,' and then I turn around, try it again, and it's like, 'No, you're not following along.' It will agree with you and tell you every idea you have is a good idea.
Yes, that's the worst part of it. I always have to tell my custom GPTs, 'I need you to dissent very hard,' and it does it right—it becomes the devil's advocate. But every now and again, it's like, 'Well, what if I had a good idea?' We're catching up with Glenn Sanford, the founder of EXP Realty, talking about AI and the future of its integration into real estate here from the Miami Beach Convention Center on KGCI Real Estate on Air. Let me know if I don't want to blow your ears out on those—they can get a little loud. Sir, you've built one of the most scalable real estate platforms on the planet. How is AI now helping you scale this whole thing smarter, not just faster?
Yeah, the reality is that there are a lot of things around EXP that used to take a lot of developers, and developers cost a lot of money—125, 150, 200,000 a year for a good developer, maybe 250 or 300,000 a year. They only have so much capacity to build in the traditional sense. Even if you have 100 people at $200,000 a year, that's a big overhead. So now, one of our big pain points was our community. We lost Workplace by Meta and went over to Slack. I was suspect of Slack, but I wanted to give it the support it deserved as a new platform. It turned out it doesn't match up with the persona of real estate agents. It's fine for staff and developers, but not good for community.
