does AI know more about real estate than we do?

The rennie podcast is about the real estate market and the people connected by it. Tune in for monthly discussions making sense of the latest market data.

EPISODE #82: DOES AI KNOW MORE ABOUT REAL ESTATE THAN WE DO?

Join Ryan Berlin, Head Economist and VP Intelligence, and Darrell Koopmans, VP Technology, as they look at how artificial intelligence is showing up across real estate and the wider economy. They discuss what AI actually is, how consumers and advisors are using it, and where it adds value in forecasting, analysis, and workflow automation. They also explore the limits of black box models, why human insight still matters, and how rennie is integrating AI in a thoughtful and practical way.

Featured guests:
Ryan Berlin, Head Economist and Vice President of Intelligence
Darrell Koppmans, Vice President of Technology

Thank you for tuning in to The rennie Podcast, where we share our passion for homes, housing, community, and cities. We hope today’s episode sparked the same curiosity in you that drives us every day. If you enjoyed the conversation, don’t forget to subscribe and follow us on your favorite podcast platform. And if you have a moment, we’d love for you to leave a review—it helps others discover the insights, analysis, and perspective we bring from the rennie Intelligence division. For the latest market updates, be sure to register at rennie.com.

Have a real estate question? You can either email us at intel@rennie.com, or leave a voicemail at speakpipe.com/therenniepodcast, and we’ll try to respond in our next episode.

We’d love to answer your real estate questions. Email us at intel@rennie.com or leave a voicemail, and we’ll try to respond in future episodes. 

Transcript

Introduction:
90% of employees are using AI today, whether their employers are paying for it or not.
AI is really, you know, computers performing tasks.
I mean, it only took us about a day to think that we're all losing our jobs.
But it was like, it was an incredible, incredible experience.
And it feels like magic, right?
Yeah, it does.
It feels like magic, and I think it's important to know that AI isn't thinking.
They think they can increase their revenue by 75% without increasing headcount.
Mm. Yeah, it's just gonna impact jobs, right? I think there's a real concern there. 

Ryan Berlin: Welcome to The rennie Podcast. I'm Ryan Berlin, rennie's Head Economist and Vice President of Intelligence. And we're recording this episode from our office in Whistler, which we don't do for just any episode. Today we're not going to talk about real estate trends and the housing market per se, but instead, about a topic that has perhaps superseded real estate in the consciousness of British Colombians. And I'm not talking about the Toronto Blue Jays, I'm talking about artificial intelligence. So I want to share a few data points before we get started here. 90% of employees are using AI today, whether their employers are paying for it or not. 82% of consumers, people who are in the market for housing, either selling or buying, are using it to access real estate data and understand the market better. However, a separate study by MIT revealed that there's a 95% failure rate for enterprise AI solutions. So, in other words, 95% of so-called gen AI pilots that companies have funded and tried to implement have flopped, they haven't worked. So we have this ubiquity of usage, but a lack of transformative impact, which is why some people look at AI and ask the question, "Are we in a bubble?" So we'll touch on that episode. We're also gonna talk about, from the top, what AI actually is, how it's showing up in real estate, how consumers are using it, and what it means for the real estate industry. I obviously cannot do this alone. This is outside of my area of expertise. So, with me today is a good friend, somebody that I've worked with for the past 10, eight, nine years?

Darrell Koopmans: Let's call it 10.

Ryan Berlin: Call it 10. 10 years. Uh, Darrell Koopmans, who is the Vice President of Technology at rennie, the so-called AI czar. He wants us to call him Czar around the office.

Darrell Koopmans: [laughs]

Ryan Berlin: I just call him DK. So DK, welcome to the podcast.

Darrell Koopmans: Thank you for having me.

Ryan Berlin: Before we get started, why don't you tell our listeners and our viewers

Ryan Berlin: what you do at rennie, how you spend your time, where you make an impact?

Darrell Koopmans: Yeah, I've been at rennie for eight years. Uh, our division, uh, supports all things technology throughout the business. It's from all the corporate IT support, uh, network infrastructure at all our locations, hardware and software procurement. We deal with rennie.com, our website. It's a custom developed, uh, platform. Uh, we also develop apps for our realtors, uh, through a platform called rdesk that we created. And we support the rCatalyst, uh, sales system as well, which is for the whole developer, uh, side of the business, uh, for all the way from lead to contract. Um, we track, um, our, our purchasers through there. And as well as our data architecture, which supports your division quite heavily, which consists of a data lake, um, about 70 to 75, uh, data pipelines that we feed into the system that we're monitoring daily and, and making sure there's [laughs] some solid uptime there. And of course, over the past year, uh, really looking at AI, how it impacts our industry, our business, and, uh, and how we infuse that into, into our workload.

Ryan Berlin: Awesome. Yeah, I mean, what you guys do is absolutely critical to the intelligence function at rennie, and I appreciate you guys every day.

Darrell Koopmans: [laughs] 

Ryan Berlin: So let, let's turn to the topic of AI more fulsomely here. And I, I, I wanna start with what seems like a very, to you, it'd be a very basic question, probably one that doesn't need to be asked, but I wanna ask it and I'll, and I'll get to it in a moment. And I guess I just wanna step back for a minute and say that, like, I don't know how many people I speak for when I say this. For me, my experience, my conscious experience in interfacing with AI, if you will, I feel like has a start date. Like, it began in December of 2022. We were in the office and a colleague called me over to his desk and a few other people and said, "Hey, look at this," and pulled up something called ChatGPT, and, you know, asked it to write something silly, like something totally asinine, like, uh, a haiku about, uh, a Pokemon character making peanut butter and jelly sandwiches. Like, I can't remember, but it was something along those lines. It had no real value, but it was amazing.

Darrell Koopmans: It was amazing.

Ryan Berlin: It was, it was like, "Wow, how did that do that?" And then the thing is, now that seems so, uh, just, just par for the course. Like, yeah, of, of course you can create that. But at that time, like prior to that moment, in, in my mind, you couldn't do that. And then he went over to this DALL-E app, an image generator-

Darrell Koopmans: Mm-hmm

Ryan Berlin: ... and created in a Van Gogh style painting, uh, this image of, like, the Super Mario Brothers plowing a field, right? So again, no utility here. Just, just more eye-opening, going, "Holy smokes." I mean, it only took us about a day to think that we're all losing our jobs. [laughs]

Darrell Koopmans: [laughs]

Ryan Berlin: But it was like, it was an incredible, incredible experience. Then again, it feels like AI has just been there in the background for a long time, generating those Netflix recommendations, supporting dynamic pricing for flights and concert tickets and sporting events, and also powering those chatbots that you see on websites. I don't know if that's AI, but my guess is that it's been in play for some time.So based on the moment, I'd love for you to talk a bit about sort of the evolution of AI, like how we got to this point. But, but just to start, to go to that, that simple question that I alluded to earlier-

Darrell Koopmans: Mm-hmm

Ryan Berlin:... what is AI?

Darrell Koopmans: [laughs]

Ryan Berlin: Succinctly.

Darrell Koopmans: For sure, succinctly. Uh, you're absolutely right, Ryan. AI has been around us for some time, definitely in things like Net- Netflix recommendations, Google Maps, like, routing you to a different road because it knows that there's traffic-

Ryan Berlin: Hmm

Darrell Koopmans: ... uh, up ahead. And really, you know, AI is this, this big broad bucket. If you think about it as the world, and then you have a continent and a country and a city or a state, this kinda nesting, uh, uh, scenario where, uh, you have all these different layers of AI i- inside. Um, so, you know, really since the '50s, um, is probably, like, the most common referred to as when AI started.

Ryan Berlin: The '50s. So, like, '70, some odd years ago. [laughs]

Darrell Koopmans: [laughs] Yeah. And you have, uh, Arthur Samuel, 1952. He's creating a program to, uh, learn how to play checkers. The founder of IBM, uh, Thomas J. Watson Sr., at the time, uh, had also said that based on the demonstration that he saw from Arthur, uh, of this program, that his stock price was gonna jump 15 points, and it did. And so these things have been there in e- in existence for, um, for quite some time. And, and, uh, if you go on Wikipedia, you can see, uh, you know, there's a, there's a timeline of AI, right? And there, there was a AI winter at one point, and, uh, you know, where, where there's no inno- innovation happening. Now, to your point, in, in December, [clears throat] it really opened everyone's eyes to, you know, what's possible. And it also opened everyone's eyes to, oh, actually, this stuff has been around us in some way, shape, or form. AI is really, you know, computers performing tasks that are typically done by humans. Things like understanding language, recognizing images, making decisions, and now generating content through things like ChatGPT, Gemini, generative AI.

Ryan Berlin: And that seems like the new ... I mean, that's what's really sort of brought it front and center because now individual users can access tools that are leveraging generative AI. So they're, they're creating something out of nothing seemingly.

Darrell Koopmans: Exactly.

Ryan Berlin: Um, and before we, we, we didn't have that capability at the, like, individual level.

Darrell Koopmans: And it feels like magic, right?

Ryan Berlin: Yeah, it does.

Darrell Koopmans: It feels like magic. And I think it's important to know that AI isn't thinking.

Ryan Berlin: Right.

Darrell Koopmans: Right? It's really recognizing, uh, patterns in the data that it's trained on, right? So if you think about these, uh, mass models like, uh, like ChatGPT or Gemini, they're looking at large, vast amounts of data, whether that's text, images from the internet, whatever it may be, and then they're able to recognize patterns. And when a user asks for something, magically it's able to produce that relatively well, right?

Ryan Berlin: And so what you described there, that would be like an example of machine learning.

Darrell Koopmans: For sure. Machine learning, AI systems learning from patterns and data, right? And there, and there's a few, uh, nuances to that, and there's different ways that a machine could learn. There's supervised lear- learning where you're just giving it the answer key. You're telling it-

Ryan Berlin: So a human is?

Darrell Koopmans: Yes.

Ryan Berlin: Like a human?

Darrell Koopmans: A human is saying, [clears throat] "Here's your answer key. It's either spam or not spam." Right? And-

Ryan Berlin: Right

Darrell Koopmans: ... and it's able to categorize the data that way. Unsupervised learning is where you just give it all the data and let it figure it out.

Ryan Berlin: How does it figure it out if it doesn't have something to lean, lean on, like-

Darrell Koopmans: Yeah

Ryan Berlin: ... like an answer key to know that ... Is it more that it just categorizes, and then after the fact-

Darrell Koopmans: Exactly

Ryan Berlin:... a human says, "Yeah, you got it right here."

Darrell Koopmans: It's looking for groupings, right? In the data, right? It doesn't necessarily know what it means exactly, right? But it's giving, it's giving the groupings and you're able to define what those groupings mean at a deeper level later. There's also reinforcement learning, and you'll see this in ChatGPT as well, uh, where you can, like, give it a thumbs up or a thumbs down if you'd like that response or not. So it's, the user is giving feedback to the system of, "Yes, you did a good job here," or, "No, you did not g- do a good job here." And it might prompt you as well to say, like, what should be different, right, next time I give you this response. And so, uh, you're helping the system learn along with the user feedback.

Ryan Berlin: And I think that's something that it took me a while to get my head around was that I thought of AI as sort of this, like, independent technology detached from everything else that just sort of created an almost like, in air quotes, "thought" and made predictions. But ultimately, it requires a lot of manual input and assessments of what it's doing.

Darrell Koopmans: For sure. It still needs data scientists. It still needs machine learning experts to understand the data, the model. I think that's one of the biggest pitfalls, is a company that does not have their data organized in a way that can be read by a machine-

Ryan Berlin: Mm-hmm

Darrell Koopmans: ... or there's gaps in their data or errors, you're not gonna see very good results when you try and implement AI. That's where a lot of these companies coming into place-

Ryan Berlin: Right

Darrell Koopmans: ... yes, they wanna help you build an AI solution, but they really want to work with you to, uh, get your data strategy set or organized. And thankfully, we have that pretty well dialed at rennie.

Ryan Berlin: So we talked about AI as a concept. Machine learning is, um, is more of a method for achieving artificial intelligence.

Darrell Koopmans: Mm-hmm.

Ryan Berlin: Is that ... Does that make sense?

Darrell Koopmans: Yeah. For sure.

Ryan Berlin: A subset?

Darrell Koopmans: Exactly. And I think, you know, there's others like deep learning, right? 

Ryan Berlin: Mm-hmm.

Darrell Koopmans: Deep learning is where you have ... It's almost trying to imitate a human brain. Neurons talking to each other. So if you think about it has these layers of neurons along the way. Take image, uh, recognition, for example, right? So you, you give a system, uh, an image of a cat or a dog. It's looking for ...... first, first layer, first neuron. It's going through and looking at the edges, right? And what is the outline doing? Okay? Like, is it, uh, a tail is protruding over here, right? It's recognizing different things in its overall shape. Then you start to, like, dive into things, like, the next layer might be colors, right?

Ryan Berlin: Mm-hmm.

Darrell Koopmans: So, w- what colors represent, uh, are, are shown in this image. And then more, like, uh, facial features or, or shapes of, you know, of the animal itself, ears, eyes, where those things, um, start to go. And then ultimately, the last layer, it's merging them all together, and it's looking for now it has all this information about how this image is broken down, and I can safely assume that this is a cat [laughs], right?

Ryan Berlin: Right.

Darrell Koopmans: Or this is a dog, or whatever. And you're also, again, you might be feeding that, that model with, "Here's, like, 3,000 pictures of cats. Here's 3,000 pictures of dogs." And it's learning based off of all those shapes and neurons of, of what it's doing exactly.

Ryan Berlin: Right. And it's so good now, it can differentiate between dogs and pain au chocolat.

Darrell Koopmans: [laughs]. Yes.

Ryan Berlin: Do we know that ... Do we know that meme? So again, my, my knowledge of AI is not very deep. It's very surface level, but I am aware there's different types of AI. There's generative AI-

Darrell Koopmans: Mm-hmm

Ryan Berlin: ... which is, again, how it seems like a lot of people, um, that, that is their experience, their hands-on experience with it through, like, these, like, ChatGBTs and Geminis.

Darrell Koopmans: Mm-hmm.

Ryan Berlin: But then there's also discriminative AI, which is used to predict outcomes. So, maybe you could talk a bit about those and, you know, where we are today with those technologies.

Darrell Koopmans: Yeah, for sure. Uh, yeah, let's start with the discriminative side. I mean, this is, again, stuff that has been around for, for quite some time. It is really, um, the data, learning from patterns, and understanding what is in that dataset. And as new data's coming in, it's saying, "Most likely, I can categorize this as this," right? It's really probab- probability-based. So, if you think about things like, uh, market forecasting, you know, we can look at the history of where market rates are at, where employment rates are at, where home sale prices are at, right?

Ryan Berlin: Mm-hmm.

Darrell Koopmans: And we can start to layer in these, these different datasets to really help predict, you know, what, you know, what we would expect, maybe a home to sell at. And, you know, groups like Zillow have done that, you know, uh, things like Zestimates, um, you know, exist. Um, so there's a lot of, like, AVM tools out there-

Ryan Berlin: Right

Darrell Koopmans: ... and whatnot. And now what we're seeing is this, this generative AI side, which is really, you know, where, where the boom is happening, and it's really an arms race of these different companies deploying, you know, a better model and one that's more accurate or produces images faster or gives you a deeper research report, uh, maybe that you couldn't get from a different model. Some of the models code better, some, uh, than others.

Ryan Berlin: Right.

Darrell Koopmans: Um, so there's a lot of different use cases. Um, I myself use three or four different gen AI tools, uh, for different reasons and, and purposes. But yeah, really, it's, you know, it's, it's created this new opportunity for people to learn, um-

Ryan Berlin: Mm-hmm

Darrell Koopmans: ... uh, to, to try new things. Uh, my son, uh, started a chef bot. [laughs] He wanted to learn how to cook, uh, short ribs, and so, uh, he created a chef bot that speaks gen alpha language and gives him all the recipes and the-

Ryan Berlin: Right

Darrell Koopmans: ... ingredients list, and he just has to input what he has, and he spent three hours in the kitchen making ribs for the family. So, you know, for, for a 12-year-old to have, um, you know, access to some of these things, it's really, um, kinda changing the world, uh, daily. Yeah.

Ryan Berlin: Yeah, it's ... It reminds me of when [laughs], when we got the internet, and all of a sudden you had information that was accessible not through, like, Encyclopedia Britannica or the-

Darrell Koopmans: Mm-hmm

Ryan Berlin: ... Microsoft Encarta [laughs] CDs, but it was, it was just a, a much more vast database that existed, and, and you could tap it to see what was out there. And this is now seeing our kids sort of experience it in the same way. They can ask questions like they would ask, uh, an expert in whatever it is that they're, they're trying to figure out.

Darrell Koopmans: Yeah.

Ryan Berlin: And there you go. They can make meals that are tailored to dietary needs and calorie counts and all that stuff. Um, so all that's, like, really interesting and fascinating, and, and, and you can see why, you know, everybody is, you know, seemingly all in on this technology. Let's bring it to real estate, though.

Darrell Koopmans: Mm-hmm.

Ryan Berlin: You, you talked a little bit about, you know, how it does show up in real estate, but there, there's a multitude of ways. So maybe, um, you can give some examples of how companies are using it and also how we're using it at rennie.

Darrell Koopmans: Yeah, I think there's, there's multiple ways. I think a lot of the companies right now are really looking internally, so how from an operational efficiency standpoint, uh, can I enable my staff to work on, uh, higher value task items versus lower task things like-

Ryan Berlin: Mm-hmm

Darrell Koopmans: ... sending emails and booking things in Calendar. So definitely from an internal perspective, um, there's a lot of automation and capability there. We're seeing it in the building industry, from like a climate risk assessment and forecasting that way, uh, from HVAC improvements based with sensors in the wall and, and technology leverage there to make more energy-efficient homes as well. So, a lot of interesting stuff on, on that side. From the real estate industry at, at large, you know, things like, um, AVM tools, um-

Ryan Berlin: Which is-

Darrell Koopmans: Automatic-

Ryan Berlin: ... automatic valuation model. [laughs]

Darrell Koopmans: [laughs] Yeah, that's the one, automatic valuation model. So, understanding what a home could be priced at and, and valued at-

Ryan Berlin: Yeah

Darrell Koopmans: ... and what it could potentially sell for. And we've seen that, again, through Zillow Estimates and, and other groups have, have come up with their own models as well, um, outside of that. Things like home staging. So, you know, realtors are able to use, um, these generative tools to drop in their listing and have it stage the home. So if there's no furniture in the home, they can choose a theme that they like if they want a coastal theme, fill all the ... all 40 photos with, um, furniture and paintings on the wall and ... to match that style. So, things like that have really, uh, allowed individuals, realtors to kind of take things into their own hands at a cost that is, you know, $20 per month versus, you know-

Ryan Berlin: Right

Darrell Koopmans: ... paying a professional company to, to stage a home for them.

Ryan Berlin: So at rennie, how do we ... how do we look at the deployment of AI? Like, what are the considerations that we're making in applying that technology to what we do?

Darrell Koopmans: We're really looking at this in three kinda key buckets. Revenue generation. So, how do we connect our consumers with-... our realtors quicker-

Ryan Berlin: Mm-hmm

Darrell Koopmans: ... or our consumers with real estate quicker. So really an opportunity for lead generation and, and revenue generation there. Second bucket is on operational efficiency, which I mentioned. So where's that agent that's able to sit beside each staff member and give them an hour of their time back every day?

Ryan Berlin: Mm-hmm.

Darrell Koopmans: How are we able to produce more content? Or every time we add a new region, let's say, and we ingest a whole bunch of new data, are we able to analyze that data quicker, faster to get a report out, um, to our clients? Third bucket is, is really on spend. So, you know, we spend quite a significant amount on, on software-

Ryan Berlin: Mm-hmm

Darrell Koopmans:... at rennie. And so how do we leverage AI to really allow these systems, A, to talk to each other, to maximize their value, uh, in the background through different connections and AI agents. Or we might not need three or four of the pieces of software anymore-

Ryan Berlin: Right

Darrell Koopmans:... 'cause we're able to solve it with a ChatGPT, a Gemini integration, or what have you.

Ryan Berlin: Yeah, it's interesting, you talk about operational efficiency, and I think that se- sends chills down-

Darrell Koopmans: [laughs]

Ryan Berlin: ... some people's spines. Um, and we're hearing there's actually a lot in the news right now about some of the, um, larger tech companies, particularly in the US, that are looking at, um... There's some internal documents that were, that were leaked from Amazon-

Darrell Koopmans: Mm-hmm

Ryan Berlin: ... where they think they, you know, over the next, like, I think it's, you know, five to six years, they think they can increase their revenue by 75% without increasing headcount.

Darrell Koopmans: Mm-hmm.

Ryan Berlin: And some of those other, um, big five tech firms are now slashing jobs for, for a whole bunch of reasons. I mean, I think some of it has to do with the, the splurge on hiring that occurred in 2021 and 2022.

Darrell Koopmans: Mm-hmm.

Ryan Berlin: But certainly, AI is starting to play a role here where, as you said, for some of the, some of these, like, lower value or repetitive tasks, at least, you know, the technology can do as good a job or better than, than perhaps humans. The way I look at it for myself, like within intelligence is we're not using AI to reduce headcount. We're using it to increase our capacity, like our ability to work through large data sets, to forecast, you know, home sales, not just for one region, but for all markets and for sub-markets, and to be able to produce more content than we otherwise would have. So, you know, I think when I look forward, one thing that I've, I've, I've come across is, um, there are more people who are, you know, in, in recent months than, than ever before that are producing, using this deep research function in ChatGPT or whatever it's called with the other, the other bots and producing these so-called market reports. Right?

Darrell Koopmans: Mm-hmm.

Ryan Berlin: So, like, tell me what's happening in Vancouver's housing market-

Darrell Koopmans: Yeah

Ryan Berlin: ... and, and economy.

Darrell Koopmans: Yeah.

Ryan Berlin: And it's pre- it's impressive, like it is. I mean, you give it five to 10 minutes and it produces this, you know, like a research paper. The one issue I have is that when I read through those types of reports, there's a lack of transparency, a lack of understanding of what the context is for some of the data points or where the data come from.

Darrell Koopmans: Mm-hmm.

Ryan Berlin: I mean, there's citations-

Darrell Koopmans: Mm-hmm

Ryan Berlin: ... but you don't really know what it's speaking to.

Darrell Koopmans: Mm-hmm.

Ryan Berlin: And you don't always know what's been excluded and why. And I think when it comes to modeling, it's very tempting to just defer to AI. Like in my world-

Darrell Koopmans: Mm-hmm

Ryan Berlin: ... it would be so easy to just... I mean, easy in a sense to just turn-

Darrell Koopmans: Mm-hmm

Ryan Berlin: ... things over to the technology and say, "Start forecasting everything that we talk about for every market we're in." And let's start pushing that out to our clients and to, you know, consumers and that-

Darrell Koopmans: Yeah

Ryan Berlin: ... kind of thing. But that's, that's black box modeling, right? And there's not a ton of utility in forecasting the what's if you don't know the why's. And I've spent my career when I am forecasting or doing any analysis, doing it in a way that is as transparent as it can be, so that... And, and where I know the inner workings.

Darrell Koopmans: Right.

Ryan Berlin: Because when I stand in front of a client and I say, "Hey, you know, over the next couple of years, we expect that, you know, rental rates, for example-

Darrell Koopmans: Yeah

Ryan Berlin: ... are going to decline, and then we think they'll start rebounding in year three and beyond."

Darrell Koopmans: Yeah.

Ryan Berlin: Well, the very first question they ask is why? [laughs]

Darrell Koopmans: Yeah. [laughs]

Ryan Berlin: Right? So it's, it's great to be able to have the ability to forecast and to make that initial statement.

Darrell Koopmans: Yeah.

Ryan Berlin: But you kinda get stuck then when someone-

Darrell Koopmans: Sure

Ryan Berlin: ... asks you why, and you don't, you don't understand what the inputs are or-

Darrell Koopmans: Yeah

Ryan Berlin:... how they're being used.

Darrell Koopmans: Yeah.

Ryan Berlin: So to me, there is... This goes back to this idea of, um, having tacit knowledge around or subject matter expertise-

Darrell Koopmans: Mm-hmm

Ryan Berlin:... and combining that with AI. And that being where the value is as opposed to just using the-

Darrell Koopmans: Yeah

Ryan Berlin: ... technology and it replacing the knowledge that people possess.

Darrell Koopmans: Yeah. So you, you touched on a few things there. So I wanna go back to the employment remarks you made, uh, just around, yeah, is this gonna impact jobs, right? I think there's a real concern there.

Ryan Berlin: Mm-hmm.

Darrell Koopmans: Like it's, it's valid, you know, from what we've seen in our ability to use it in coding, uh, and our developer able, able to produce around 10 weeks of development time in a week and a half of, of prompting and, and guidance.

Ryan Berlin: Mm-hmm.

Darrell Koopmans: Now, to your point, this thing was going off the rails and [laughs] it was all over the map and started introducing, like, new issues or do things like that. You still need that tacit knowledge to really review the code, understand what's going on, prompt it the right way-

Ryan Berlin: Mm-hmm

Darrell Koopmans: ... right? Give it the right directions and, and guard rails so that it's not going off the rails. But in, in groups where you have relatively small teams, you know, you're able to, to actually achieve, like, quite a sig- significant amount of, of output. But you have to be cautious of those hallucinations, right? So is this deep research report I'm getting on the Vancouver market actually valid, right? Do the numbers line up? Uh, does it make sense? 'Cause it, it is pulling from 80 plus different sources. And I think...... you know, the unique thing that we're able to do at rennie because we have our whole data architecture is that we can lock it to just that framework.

Ryan Berlin: Mm-hmm.

Darrell Koopmans: Right? It doesn't need to pull in outside sources. We're feeding it the, the exact same sources that were producing the rennie Review, or the, the pre-sale snapshots and things like that.

Ryan Berlin: Mm-hmm.

Darrell Koopmans: We're just giving that to the AI, right? And so the hallucination that it can produce is a lot more guardrailed. Now, still needs your knowledge, still needs your team to-

Ryan Berlin: Mm-hmm

Darrell Koopmans: ... go through it all. Are we comfortable putting this report out? But it's now enabled you to get you 80% of the way there in a couple hours versus maybe three weeks of effort to create that report.

Ryan Berlin: Absolutely. And that's where I can see it, you know, in the future reducing the need to expand workforces. I think that's where I see, in a broad sense, a lot of the impact. And my, my concerns in the short term are that there are disruptions. And, um, you know, long term, I think the economy and society, our education system, everything will adjust. You know, we went from an agrarian society into the industrial revolution. Computers were invented. We had the PC revolution, the internet, and now AI. It's like through all of those changes, there were disruptions along the way, but our ability to produce and consume grew. The economy grew. More and more people were employed. People were paid more over time, or earned more. And so this is just another evolution. There's a speed to this that we haven't seen. [laughs]

Darrell Koopmans: Yeah. Sure.

Ryan Berlin: A speed of transformation that we haven't seen-

Darrell Koopmans: Yeah

Ryan Berlin:... in the past.

Darrell Koopmans: It's a hockey stick curve, right? The technology is advancing. The compute power is advancing like, so fast and able to support all this stuff as well, which has all its knock-on effects, right, of environmental impact, ethical issues, right? These are all very similar discussions that they were having-

Ryan Berlin: Mm-hmm

Darrell Koopmans: ... when the, the internet was born, right? All these things are real-life conversations that need to be had. Also, you, you mentioned, you know, this notion of a black box. 100% agree. There's always the why, right? Why, why are we getting this number? Why are you saying what you're saying? You know, my, my notion to our company and to our advisor base as well is, be curious, right?

Ryan Berlin: Mm-hmm.

Darrell Koopmans: We actually don't know what we're gonna get if we don't actually try it, right? So, it might be a black box. It might be something that we don't need to publish, you know, in, in the public or consumer eye. But if we don't know what its capability is or its flexibility or, yeah, understanding of our systems, what could talk to what, there's a real opportunity there to, to just be curious and, and to just try things because you actually don't really know what you're gonna get. [laughs]

Ryan Berlin: Yeah, I think that's a great, a great piece of advice. And maybe just to close off, your thoughts on in the real estate sector itself. You know, we, we, we talked about [laughs] uh, AI, you know, threatening some jobs.

Darrell Koopmans: Mm-hmm.

Ryan Berlin: If we talk about realtors, for example.

Darrell Koopmans: Mm-hmm.

Ryan Berlin: Here, here's a direct question. Can AI take the place of realtors?

Darrell Koopmans: No. [laughs]

Ryan Berlin: Why not? [laughs]

Darrell Koopmans: I'll go back to ... Y- you mentioned tacit knowledge. You know, w- we'll go back to that. I think ultimately, AI will take us to a point, right? It'll take us to, uh, uh, maybe a, a really good spot where we learn something new. Something was gained by, by us, right? Uh, I'll, I'll use an example of, um, there's a, a commercial piece of property in Squamish that I was kinda interested in le- and it was close to my house. And I know nothing about commercial real estate, even though I've been in this industry for 23 years, outside of there's commercial agents that, [laughs] that I work with in that space. You know, all of the, the, the net rates and things like that, I have a high-level understanding. So, I was curious, well, can I, you know, afford the caring costs of, of this building, right? But I don't really know all the ins and outs of, of how to calculate that. So, I used AI and had a discussion with it around this property. I gave it the, the, the feature sheet. I uploaded my home as well, uh, you know, so it could calculate kind of the whole value of the whole property 'cause it was right behind my house. And, um, it got me, like, quite a ways like along, but, uh, at, at the end of it, I was kinda stuck. I was like, "Well, I don't know how much rent-"

Ryan Berlin: [laughs]

Darrell Koopmans: ... [laughs] this place is generating. I don't know anything about this building, like it, it ... the ins and outs of it. I don't know how to purchase this building. I can't write the contract myself. I don't have that skill or capability. So, I think it'll, it'll always bring us to a point where ultimately, we'll still need that expertise, right? I still need ... Again, even though I'm reading your reports, [laughs] right, about real estate in Squamish, I don't know in five months time, like, is this gonna be a good investment for me, right?

Ryan Berlin: Mm-hmm.

Darrell Koopmans: ChatGPT is not gonna be able to tell me that because all it's doing is grabbing information that exists already and giving me an answer that I wanna hear. And so, uh, at the end of the day, you know, we still need that professional human relationship, [laughs] right? And, and I still need that expertise. And I'm not gonna go back to school and learn how to, you know, be a commercial realtor just so I can purchase this one property. But I think, again, that's where there's opportunity for our realtors as well, to understand that, you know, consumers ... You mentioned that stat earlier, 82% of-

Ryan Berlin: Mm-hmm

Darrell Koopmans: ... consumers are using AI already to find housing information. How can our realtors, advisors be well-equipped to have that conversation with their client, right? Well, this is what AI's doing, right? Just, just FYI, this is what it's doing. It's looking at past information. It doesn't know the heartbeat of the community. It doesn't know who bought that property down the street is now gonna buy, you know, the next three over so they can build a condo. You know, there's a, there's a lot of that, uh, just knowledge and relationship that you're never gonna garner from an AI system.

Ryan Berlin: Well, and also, there's just the trust that comes with human interaction that doesn't exist between humans and technology.

Darrell Koopmans: That's true.

Ryan Berlin: I don't know if we'll ever get there. And, and when you talk about realtors, there's ... They do more than just, you know, get the most value for a property you're selling, or try to get a property for the best price. There's so much more that goes into-... transactions.

Darrell Koopmans: Yeah.

Ryan Berlin: And about closing timelines and conditions of sale, where I know personally from my experience, working with a, a human who was able to work through our needs was absolutely invaluable. It wasn't just about, you know, spitting out a, or landing on a value that, that worked for everyone. There's a lot more to every transaction.

Darrell Koopmans: And, and it's that last mile, right? It's that last mile problem. But that last mile can actually take a really long time, right? 

Ryan Berlin: Yeah. Sure.

Darrell Koopmans: Which needs relationship, which needs guidance, which needs your expertise. And to finish, you know, this commercial notion that I went through, I wound up contacting, uh, Gena Belanger, a Squamish realtor, and just said, "Hey, I need to know more [laughs] about this property. Does it make sense? What are the rental rates?" And she was able to get me that info super quick. And so, it actually created that connection.

Ryan Berlin: Mm-hmm.

Darrell Koopmans: Right? If I, you know, only had Google, let's say, in the past, I might have only just stopped there, like, "You know what? Too complex, too much information."

Ryan Berlin: For sure.

Darrell Koopmans: "I'm not gonna dive into this commercial thing. I don't really get it." Generative AI has, has opened up some of those doors of, like, of new learning, new information, to at least then, you know, cause me to, to go reach out to my realtor and, and find out, [laughs] is this even feasible, right?

Ryan Berlin: Absolutely.

Darrell Koopmans: Does this make sense?

Ryan Berlin: All right. So, to wrap up, there, there, we've talked about a lot here. It's still, like, for me, it's this world of AI is, is, like, overwhelming, especially the speed at which things are evolving. It's hard to keep up. But if you had to say, like, for, for, for businesses, whether it's in real estate or outside of it, what is one thing that they can do or need to do to realize the, the, the potential of AI?

Darrell Koopmans: Hmm. Yeah, great question. I, I think it's hard to even predict what's next, and I think for businesses today, the key focus is really their data strategy. If you think about all these systems, tools, chat bots, they're all reading your data, right? And so if your data is misaligned, has gaps or errors in it, you know, your actual output, as we mentioned earlier, is not gonna be great, and you're gonna have a lot of difficulty with it. So, I do think data is really key, uh, in a- in AI. It's also an opportunity for, for businesses to [laughs] to review that data, right, and to really assess it, ensure it's the right data, that they're supposed to have personal identifying information, things like that. It's a good exercise for, for a business to go through, and then layer on the AI on top of that, right? Bring in some expertise, either in-house or outsource. We found that has gone a long way with groups that we've talked to, where our data strategy and our architecture is already set, as we're well ahead of, of where most companies sit today.

Ryan Berlin: Obviously, thanks in no small part to the expertise you brought to rennie on the AI front. Thank you for the expertise that you brought to this conversation. Couldn't have done it without you, Czar.

Darrell Koopmans: [laughs]

Ryan Berlin: Um, so thanks for joining us today.

Darrell Koopmans: Thanks for having me, Ryan. A lot of fun.

Ryan Berlin: For everybody who is watching on YouTube, uh, if you have any questions about anything we discussed today or about real estate or AI in real estate, put your questions in the comments, and we'll get back to you. If you're listening to us and not watching, head over to YouTube and do the same thing.

Thank you for joining us on The rennie Podcast. If you'd like to learn more or to subscribe to intelligence updates, go to rennie.com/intelligence.