APU Business Innovations in the Workplace Podcast

Podcast: Using Artificial Intelligence in Residential Real Estate

Podcast with Dr. Wanda CurleeProgram Director, School of Business and
Dr. Zona KosticProgram Director of Computer Science

What are the benefits of applying artificial intelligence to the residential real estate market? In this episode, Dr. Wanda Curlee interviews APU computer science Program Director Dr. Zona Kostic about the advancement of AI algorithms in real estate applications. Learn how real estate platforms are using AI to accurately assess property values, better understand customer desires and behavior, and automate processes to assist real estate transactions. Also learn how drones are being used during the pandemic to collect more information about the location of a house, the neighborhood, and nearby points of interest and how that data can help with machine learning algorithms.

Listen to the Episode:

Subscribe to Innovations in the Workplace
Apple Podcasts | Spotify | Google Podcasts

Read the Transcript:

Dr. Wanda Curlee: Welcome to the podcast. I’m your host, Wanda Curlee. Today, we are going to be chatting about artificial intelligence in residential real estate. Today, my guest is Dr. Zona Kostic. She is the program director for computer science at American Public University. Zona, welcome. Thank you for joining me.

Dr. Zona Kostic: Hello, Wanda. Thanks for having me. It’s such a pleasure to be here with you.

Dr. Wanda Curlee: Great. You have quite an impressive background. Before we get started, how would you define artificial intelligence as it applies to the residential real estate market?

Start a Real Estate Degree at American Public University.

Dr. Zona Kostic: Oh, wow. So residential real estate market hasn’t been really disturbed by AI. I’m not sure is this good or bad. People who are real estate experts, they think that this is great because they believe that real estate is a human discipline and they don’t see a space for AI. While on the other side, I do see a lot of AI algorithms that could be useful to real estate professionals to basically help them save time and to basically help them automate certain processes.

I would say that older real estate hasn’t been really disrupted as mentioned by AI. There are some companies that been working on new algorithms, and I can also mention, Zillow is definitely one of them, Redfin, REX, a lot of startup companies. So it’s actually getting there.

This would be actually really important to mention that AI is not there to replace human expert. AI is there to help human expert coming up with a decision. So there are so many processes that AI can just take over and generate a set of results that human expert will pick up, and basically on these results they will end up deciding on a next step. So AI has a really interesting role within the real estate.

Dr. Wanda Curlee: Yeah, I had always wondered if Zillow and Redfin and all those platforms were AI driven. I assumed they were. I know my son bought his last house through Zillow, so that’s very interesting.

Dr. Zona Kostic: Yeah. Well, a lot of AI over there. First of all, Zillow is very famous by their own Zestimate value, which is completely AI driven. They run a major competition on platform called Kaggle and it was the competition that gathered 4,000 different teams for a price of $1.2 million. And truth to be told, that competition inspired me to become part of real estate world.

So after the competition, I would say that Zestimate value is definitely better. It is closer to a real value to the real sold value of the properties. It is also on the other side, it is really hard to 100% come up with an estimated value of every property because the most, the hardest possible challenge, as part of the real estate world and AI, is how to model human behavior, right? We don’t really know what a seller is interested in doing and what buyers might be interested in doing. So that’s really hard to estimate, but figures are getting there.

[Podcast: The Complexities of Processing Big Data Using Artificial Intelligence]

And unlike Zillow there, as I mentioned, other players such as Redfin or REX or Torii. All of them have different approaches. I spoke to a lot of them. I cannot really talk about machine learning algorithms they’ve been using nor any sort of future plans because I signed NDAs. But I can say that they are really fascinating in terms of results, as well as models, as well as how they see the transformation of real estate.

Dr. Wanda Curlee: Zona, so that’s quite interesting. Can you tell me what is it that AI does for the real estate market?

Dr. Zona Kostic: Oh, excellent question. Because it will help real estate professionals gaining trust in AI models. AI models, as I mentioned, they’re here to help professionals. Basically, all the roles within the real estate domain can benefit from AI. When I say all roles, I mean, buyers, sellers, agents, brokers, new construction developers. So, basically, the point of all the machine learning algorithms and AI algorithms these days is to automate certain processes that are considered tedious processes, it takes so much time from a real estate expert in order to come up with an answer.

Very simple example of it is brokers tend to spend two to three, or maybe even three to four hours every morning, browsing different units for their clients. We have AI that can help automating this process, giving the suggestion saying, “Okay, I know that your client number one, might be interested in this type of property. Here are the similar listings.” And then agent will come up with a final decision and say, “Oh yes, this one is interesting. Send it to my client number one,” or it’s just going to disregard it and say, “None of these are actually according to the taste of my client.”

So instead of browsing, they can wake up early in the morning and just come up with the answers that AI automated, AI generated, and then decide based on it. With every new decision, AI will be able to learn about the client as well as something about the real estate agent. So it’s there to basically compensate for all of these tedious processes, as I mentioned, but still it’s not going to decide without a human in the loop. So this is just one of the examples.

Dr. Wanda Curlee: So it allows the human to do the value add.

Dr. Zona Kostic: Every time. This is basically why I was interested in this domain is that it doesn’t exclude domain expert. It actually depends on domain expert. And there is this collaboration between AI as well as an expert in this domain that is grading this entire process, magical. Because AI can learn from domain experts, but domain expert can also save some time and use time to focus on a client rather than browsing units, for example.

Dr. Wanda Curlee: Interesting, interesting. You mentioned that AI is not really in the residential real estate market, not yet, but what tools do you envision that AI bringing into the residential real estate market?

Dr. Zona Kostic: So recommendations, definitely recommendations. Algorithms are going to be first because these types of algorithms will save time. And they’re also multiple roles within the real estate domain would be able to benefit from recommendations.

Buyers can benefit from its sellers, also brokers, agents. I also mention new construction, building. In new construction, everything is about what is going to be interesting in the future. So in order for me as a real estate developer to actually have an understanding what kind of building I should come up with.

And when I say what kind of building, I mean, what kind of units I should have crafted in my building. I need to know which market is going to be hungry for what type of unit. So if I’m working on a new construction project right now, I want to make sure that all of my single bedroom, or two bedroom units, or studios are going to be interested for that market two years down the road.

So it is a recommendation, but this type of algorithm is also focusing on predictions. Predictions, supply and demand, as well as of pricing. Usually pricing algorithms is something that it has been really interesting within the real estate domain. We would really like to know the price movements and the price changes over time. I could say that AI algorithm started to become important as part of the real estate domain with pricing algorithms.

But then we have different needs. Buyers have different needs. Sellers have different needs. And based on these sort of needs, we came up with new algorithms, let’s put it this way. Besides recommendation and pricing and modeling, I’m pretty sure that real estate experts would like to know prospect clients. There is also a domain of assessment. So comparables are also recommendations, in terms of comparables, is going to be interesting, especially because of so many abatements that are taking place these days. I’m not even sure if the general audience is aware of the process of abatement, but basically abatement is really tied to taxes. So all homeowners have an option to abate, to basically negotiate their own taxes based on comparable units, based on similar units on the market.

The biggest problem in this domain is that evaluators, assessors, they are looking for similar units within the same neighborhood, but sometimes similar unit happens outside of the zip code. So we cannot only focus on typical metrics that have been used so far that are basically enclosed within the same zip code in the same neighborhood. There are spatial dependencies that we can put in our algorithms, but there are so many data and so many units that are outside of our neighborhood and outside of our zip code.

So in order to find similar units, especially that are outside of the location of interests, it takes a lot of time. So in any of these scenarios, AI could be there to process an enormous amount of data and to give a suggestion and still, at the end, human expert is going to come up with the final decision.

Dr. Wanda Curlee: That’s quite interesting because I just recently sold a 150-year-old home up in the Northeast and the real estate agent had a hard time trying to put what price to put on it because there wasn’t anything comparable in our area.

Dr. Zona Kostic: Yeah. But also it depends what market might be interested in. Basically, this type of units are really hard to put in the algorithm because they’re really special. They’re unique. We actually call them outliers, so they have a special group of people and buyers who might be interested in it. This type of unit cannot be easily modeled. We can use them to learn and maybe to benefit in the future, but the outliers are really hard to model, although it’s not impossible, but it is hard.

Dr. Wanda Curlee: So let’s switch to the consumer’s perspective. What tools can the consumer use when they’re trying to sell their home?

Dr. Zona Kostic: So how I should answer this question without advertising any of the companies? This is going to be really hard. This is going to be hard. I’m just looking for, at least if I’m going to advertise, I’m looking for companies that my students create excitement and feel good about it.

So I would say that there are many applications, let’s put it that way. And there are many companies who provide these sort of services, like free services, to look for a home as well as to look for a potential buyer.

Also, these applications are really easy to access using cell phones, using iPads, as well as using desktops. So whatever the user might be interested in and whatever the user might feel comfortable with, it will be very also easy to find the application that suits that need. So if the user is a buyer, for example, and they’re interested in a house at a specific location, and they’re interested in estimating a price point of it, they could easily find the app that will help them.

I’m just trying to actually stay away without mentioning, okay, you can go to Zillow and double check prices and that will be unfair because Redfin is also doing price estimates as well as National Association of Realtors, as well as REX, so there are many companies that are offering services.

Also, there are companies that are offering services without the agent where buyer and seller can actually talk directly and the company and the app will actually mediate the entire process, will provide guidance in terms of lawyers, what is the next step? So they are serving the role of the agent and they are here basically to connect buyers and sellers. So depends on the needs. I would start with Google first, try to express need. And then probably, some of the applications will be there to actually help.

Dr. Wanda Curlee: I would imagine, yes. Continuing with that, will it eventually be the real estate agents are not needed for selling your home?

Dr. Zona Kostic: I wouldn’t really say that it. Is too early for it. It is too early. First of all, every transaction process is unique. We can use artificial intelligence to model something that’s been general. It is part of, how can I put this? Of every single, every, every broker, every agent performs exactly the same steps in order to look for a client, so we can generalize that. But every transaction process is unique.

First of all, we have two sides which are people. And I said it’s not really possible to model human behavior. And then second, we have a house. House, apartment, condo, whatever, that’s in the middle of this process, it has attached history to it. History of other transactions, maybe title, maybe deeds. So that is a randomness. It is not really possible to model it. And that’s why I see human expert being really important part of it still.

It is possible to use the application that is going to connect buyers and sellers. And then maybe it is going to come up with the suggestions in order to proceed, use this set steps. But I believe when it comes to the decision, we still need human experts. So we’re not there yet, which is a good news.

Dr. Wanda Curlee: Yes, that’s good. How will AI though, make it easier in the future for both the buyer and the seller?

Dr. Zona Kostic: That’s a very good question. That’s very good question. We usually like to say, “Oh, yeah, AI will help with transparency,” Right? By bringing information to both sides. So I believe that still both buyers and sellers are not educated enough. They don’t understand. That’s why they’re asking agents to help them. This is also, you know, has both sides to it. We are also asking lawyers to help us because we are not domain experts and we don’t have enough time to actually educate ourselves.

So I would say that AI can definitely help out bringing the information. Not really the information like I’m going to go to Google and text and retrieve information, but coming up with the summary of, “This is what you can expect the market to behave in the next three months. We are 90% confident that these are going to be changes.” And then based on that insight, buyers or sellers could come up with decisions.

Dr. Wanda Curlee: We are back. And we are speaking to Dr. Zona Kostic about residential real estate and AI. And continuing with that, how do you see AI tools evolving in the residential market to help brokers, buyers, those that want to build, build, et cetera, et cetera?

Dr. Zona Kostic: Yes. Well, there are some fascinating applications already. So AI domain, it’s huge. And when somebody says that they are AI experts, there is the follow-up question like, “Which domain or subdomain of the portion of AI?” So AI is already huge.

There are all of these different subdomains of AI are contributing to the real estate market. Either is a text processing. We can text process remarks and figure it out, is this remark positive or negative, to understand how it’s going to influence the decision of someone who might be interested in seeing this unit?

We have image processing. Image processing can tell me a lot about the condition of the unit, as well as any recent change took place. So I can evaluate it. In terms of images, there is a really fascinating project that I’m considering, starting from September, which is using drones and using drone images. Given that we are at the times of COVID. How we can help evaluators, how we can help assessors estimating the price of a unit, or actually evaluating the house in order to come up with an accurate value for the sake of estimating taxes?

Usually when there is some changes took place, evaluators needs to go and observe. Right now, they cannot do it because of the new rules imposed. So how we are going to observe a house? Well, we can send drone. So we are sending drones and guess what. We have a lot of data, which every sort of AI scientist will be very happy with. Not only do we have a lot of images, we have a lot of videos.

Now, we can process the data to understand something about the house, but also to understand something about the neighborhood, about the points of interest. So we have these multiple views of inside and outside of every house and not to mention that that can evolve into many different directions.

Dr. Wanda Curlee: It’s interesting that you said drones because when we sold our house, somebody was hired to come in and bring their drone and take videos of our house. And it was fascinating. I had never seen that done before, but it makes sense, especially in times of COVID.

Dr. Zona Kostic: Yeah, we also have virtual walkthroughs right now because we cannot go and hold open houses anymore. So we have virtual walkthroughs. We have drones that are observing the house from the outside, even better than we used to do it before because when we go and check the house, we’re not taking the ladders and going on top of the roof to inspect it. Right now, we can do some sort of virtual inspection, which can help both buyers and sellers expediting the process while raising the level of trust into specific house.

Also, on the other side, we have virtual walk-throughs that are helping people seeing the houses. There are so many decisions that are taking place based on virtual walkthroughs right now. I was, I must say, quite shocked and fascinated and at the same time that buyers are pulling the trigger and they’re buying the houses by just observing it online. But if they live in some other country or if they are not capable of going and seeing the house due to the sake of COVID, virtual walkthroughs can help it. And AI scientists are really excited about it because they have a lot of data.

Dr. Wanda Curlee: Yes, I know of three people that have bought their house totally based on a virtual walkthrough and the drone videos. So to me, that’s just amazing. I would never do that. I don’t feel comfortable with that, but I can see young people that are very comfortable. My son did it, totally based on a virtual walkthrough.

Do you see that actually maturing in the future, the virtual walkthrough, along with the drones to help people feel even more comfortable with understanding what the house can do for them?

Dr. Zona Kostic: Well, definitely, especially what you just mentioned, right? I was not sure. But then you also brought this idea of, oh, okay. I even have an example of such purchase in my house so that made me think that new generations are probably more comfortable with it. They are comfortable with applications. They base their days, their every single step—work, life, even finding a partner right now on the cell phone app. There is a huge trust put in cell phone and maybe users are not aware of that, that all of these applications are heavily AI-driven.

And if we ask users, do you trust AI? “No,” because of so many bad examples that we are reading in the news about. But if I asked, “How can you trust the application that is matching you with someone else that you’re interested in sharing your life with?” “Oh, because they know.” Well, this is also AI decision. This decision to proceed, it’s going to be yours, but the recommendation is AI based. So probably, given the situation with COVID, given this new normal that we get used to it right now, I believe that there will be many different applications. It’s just going to help mediate the process.

And it is not about the application. It is not about COVID only. It is not about this new normal. It is about us who are impatient, who live in this world that’s just constantly changes and we don’t have time to even sit and think and talk and negotiate anymore. We need to come up with decision and move on. And coming up with decision means, okay, let me have this app helping me with a couple of clicks so I can just be done with it and move on. I think this is also one of the reasons why are we proceeding this way as we are.

Dr. Wanda Curlee: I’ve done some research in AI and I always wonder, I’m not a coder. I don’t know a one from a zero. I understand AI. I know how to work with it. Many times information will come into it, then you have the black box where it does its magic, so to speak, and then something is spit out. Who has the responsibility, in your opinion, to make sure that what is spit out is correct so that the machine can learn?

Dr. Zona Kostic: I would say the responsibility is on the company that actually offers black box algorithms. That’s the danger. Black box algorithms can serve multiple purposes. They’re good for education. They’re great if we ask question and we get the answer and the answer says, “Okay, this is what algorithm suggested, but do not take it for granted.”

Problem is, right now, that everything is for sale, right? So we want to offer this black box algorithm as a promise. You just click, use me as a magic box, and I’m going to give you the answer. So the responsibility starts with a company who offers this sort of product.

We can also talk about the responsibility of the user who trusts in this product. But the company is aware of the fact that the user lacks knowledge, and we are actually selling this product. When I say we, I said, the companies behind the black box algorithms are selling products, being very well aware of the lack of knowledge that users have. We tend to shift. We always tend to put the responsibility on end users. You have a market, you have an option to choose between different projects. It is your fault that you didn’t read fine print. True, from one side.

But from the other side, we are constantly playing with the lack of knowledge and we are aware of it. And we are doing it a lot of reasons are financial reasons. There are examples in education, for example, when we don’t really want to put too much pressure on our end users, but we want to just give them the outcome, but the outcome is not going to hurt the next set of decisions that they are going to perform.

Dr. Wanda Curlee: So in that same vein, what should the buyer and seller and the real estate agent need to worry about when dealing with AI? Maybe not right now, but in the future, what should they be aware of?

Dr. Zona Kostic: Well, the problem is that they are not aware that they’re dealing with AI.

Dr. Wanda Curlee: Many people don’t realize AI’s all around us.

Dr. Zona Kostic: Yes. So we don’t even know who we are here, right? Who is actually real? So every single time they use the application, the application is going to be boosted or backed with AI decisions. Probably multiple AI models are supporting processes. One is supporting price modeling. The other one is supporting recommendation. You liked this house two weeks ago, here are the similar houses.

So good thing with real estate is that the decision is human-based, right? So the decision is not going to be on the side of the AI algorithm. AI is going to give suggestions, but the decision to proceed and accept, and maybe see something that AI suggested is always going to be on the user.

So this is why I said that that is a really nice middle ground. It is a really nice trade off. And that’s why I like this domain of real estate that uses, and I hope it will benefit from AI, just for the sake of saving time. But the ultimately decision will never be on AI. It will always be on human.

Dr. Wanda Curlee: Excellent, excellent. So in your mind, we don’t all have a crystal ball, we don’t know what the future holds, but how do you see the future with AI and residential real estate market evolving maybe in five, 10, 15 years.

Dr. Zona Kostic: Definitely getting there and I’m really happy to say that, because the level of trust improved with both buyers, sellers, mediators, even with a real estate development, real estate residential development. I see it present more and more. COVID inspired it. We started talking to brokerages, we started introducing AI as something that’s not harmful. And most importantly, it’s something that is not going to replace human role. So as long as we have trust in such algorithms, I can definitely see a room for improvement.

Also, one more signal to it is a presence of a lot of companies. If we just take a look at it, there’s so many startup companies that are prop tech companies, they are within the real estate domain, not necessarily interested in commercial real estate because commercial real estate, I mean, first of all, it’s a completely different domain. I’m not really familiar with it, but it’s a game of numbers. And is usually whenever we have a game of numbers, it applies the rules of stock market. So whatever was part of a time serious modeling that was present in stock market, it was also easily applied to commercial real estate.

But residential real estate is a different domain. And by different, I also want to say it’s more interesting for AI experts because we can extract knowledge from so many different parts, such as, we can focus on processing remarks and text and how agent is describing unit. Based on that description, we can understand attraction. Based on the images, we can understand how many people will click on this unit, put it among the favorites because they cannot come in and visit it in person anymore. Based on, as we mentioned, drones, we can understand something about a house, observing it from a different point of view, something that we were not able to do before.

So as soon as we bring all of these new applications, it’s not for the sake of applications, but it’s for the sake of understanding something, like gaining new knowledge, the level of trust will raise. And I’m really hoping that we will see more applications to basically come up with the shorter question, we will see more applications and more and more startup companies are emerging for that reason.

Dr. Wanda Curlee: Excellent. That’s great to hear. Zona, thank you very much for joining me today on this exciting topic of AI in residential real estate. I just see the market exploding with AI. Do you have any last words you would like to leave our listeners?

Dr. Zona Kostic: Trust AI, but also trust your decisions more than anybody else.

Dr. Wanda Curlee: That’s what I tell everybody too. Always verify, yes. And thank you to our listeners for joining us today. We have some exciting podcasts coming up, so make sure you stay tuned and

Dr. Wanda Curlee is a Program Director at American Public University. She has over 30 years of consulting and project management experience and has worked at several Fortune 500 companies. Dr. Curlee has a Doctor of Management in Organizational Leadership from the University of Phoenix, a MBA in Technology Management from the University of Phoenix, and a M.A. and a B.A. in Spanish Studies from the University of Kentucky. She has published numerous articles and several books on project management.

Comments are closed.