Futurecast: A (practical) AI revolution coming for multifamily?

Yolanda Muchnik:
Alright. So Chris, welcome to the show.

Chris Snyder:
Thank you. I'm glad to be here.

Yolanda Muchnik:
Alright, so let's kick this off and level set a little bit. I know the term AI is used a bunch right now, and I have found that sometimes people use it interchangeably with other concepts like machine learning for instance. So, in your words, can you share with your audience what is AI? What is machine learning? What's the difference between them?

Chris Snyder:
Sure. So AI is a commonly, you might even say misused term, but it's okay because what often happens is AI is coupled with machine learning or ML as it's often known as and so there winds up being some very blurred lines along the way. So even myself in the technology area where I know the difference, I'll just say AI for simplicity. So I would say that as far as worrying about getting it wrong, don't worry that anybody's going to be too condescending about that. But there are actually differences that are important to note. So AI generally is much more human-like, so it is something that could perceive its environments and interact with them in some way. A prime example, the one that we all know, Siri or Alexa, certainly one of the two of them, they listen to you, you talk to them with regular questions, you're not shortening it like you would say a Google search. You interact with them with natural language and they respond with you in natural language.

You can have AI without machine learning. So machine learning is taking large amounts of data, doing all sorts of analysis and providing a result and sometimes you take the results of that and feed it back into the system and get better at the process. So an example of machine learning without AI is a stock investment application. So there are applications that are within the stock investment realm where it takes into account the history of trading for every individual stock, but also the political climate, any changes within industry, any regulatory changes, weather, time of year, all of these things and helps to derive a good proposal for buying or selling stocks. That doesn't need a human-like interface. It could be all technical when it comes to the results. That's machine learning.

And if you want an example of AI without machine learning, if you just ask Siri to turn on the lights, say Siri turn on the kitchen lights or Alexa turn on the kitchen lights, I hope somebody listening has Alexa and their kitchen lights just turned on and they're mad at me, but you can have Alexa do very simple tasks that don't involve a whole ton of data. So AI can exist without machine learning. Machine learning can exist without artificial intelligence, but there is a third category called deep learning, which is where the growth is and where the excitement is. It's also where Hollywood has been for years. Deep learning is a form of machine learning, but you don't give it the rules.

So an example of this would be with machine learning, if you wanted to build an application that could tell the difference between a hawk and a falcon, and I get both out by my house and I never know which one it is, but there are differences, if you sat in there and you programmed in all of the differences like the color of their cheeks, the stripes on their chest, their wing tips, if you programmed all of that into the computer into the application and it could give you a result, that's machine learning. But if you just fed it a whole bunch of pictures and said this is a hawk, this is a falcon, and this is none of the above, and you just did that with a whole ton of pictures and it figured out the differences and it wrote its own rules, that's deep learning. And it's also known as unsupervised learning because I'm not really giving it the rules, I'm letting it learn the rules.

Yolanda Muchnik:
Interesting. I had never even heard that term deep learning before. Thank you for clarifying. Thanks for sharing.

Chris Snyder:
Absolutely.

Yolanda Muchnik:
Now, you've been in the industry for a while now and I'm curious where do you see AI most commonly applied in multifamily today? And what are some of the benefits that operators are deriving from its applications? Maybe you could also touch on some of the most interesting or most exciting applications you've seen in multifamily so far.

Chris Snyder:
So the main area where you'll see it is chatbots. So I would imagine, especially if you've had somebody that's been in property management for a while and you went to them and you said, "Do you ever get redundant or dumb questions?" They would probably pause before they could answer before they said, "Yeah, I just got one more." So questions like what are your office hours or what day is my rent due, do I have a package or anything? A chatbot can answer these with a high degree of accuracy, and there are chatbots out there that are done from a marketing standpoint like on the open web and chatbots that are on internal portals. So the question of what are your office hours would be an example of something that could be on both of them.

I would say that chatbots are certainly the most common one that I see within the industry. So far, I think when it comes to general AI applications, I think PropTech is evolving and I think the neat thing about it has an extremely exciting roadmap, extremely exciting future, but I think it's in somewhat of the early stages right now.

Yolanda Muchnik:
Got it. And you actually just touched on the future. So the pace of innovation with AI is extremely fast I've seen. Where do you see the multifamily industry heading with AI in the future, but even more importantly, what has to happen in order for us to get there?

Chris Snyder:
Sure. So right now, in terms of practical implementation it's going to continue down the line that it is because the concept from property management is what is the return on investment, so for instance, I talked about chatbots. They're great, they make people's lives simpler, but do they save money? You're obviously going to be paying money for these. Can I reduce headcount? Can I share a property manager between multiple properties? If I take this burden off of them, what can I do? So the driving factor for this is not surprisingly going to be return on investment. And as that becomes adopted, that will bring us up to the next step and the next step. So everything is going to be driven more from business value than capabilities for at least a while, but over time it will switch from operational and I would call the chatbots operational to tactical and into strategic.

So where I could really see it growing is in predictive behavior, that's where we're going to get the best value. So you want to know who's going to move out of your complex and when? Well, you don't know that very well necessarily unless you know the people or they've been complaining or they've been talking about it. And so things like artificial intelligence engines that can determine anything from vacancies, cost recovery, utility, expenses, marketing, what is it that's going to bring people to our door? What's going to help us to fill it out? Even ultimately getting to the point to where that question is about is there a return on investment? That could be answered by an artificial intelligence engine, and so that's when you have shifted from operational, which is where it really is right now and where it's going to be for a little while to tactical, how do we achieve this, to strategic what's going to happen in the future, what do we need to be doing now for our next phases?

Yolanda Muchnik:
Got it. So as with any new technology, I think success ultimately hinges on successful implementation. What are some challenges or roadblocks that you've seen operators face when it comes to adopting AI for newer use cases?

Chris Snyder:
Sure. So it's going to be the same that I've encountered, which is there is still a gap between AI and humans and sometimes that gap doesn't look big on paper, but it can change the net effect of everything. So an example, and I hate to keep going back to chatbots, but I think everybody knows about chatbots and has at least had to interact with them. Let's just say that you had one that's at 95% effectiveness. That may sound excellent, and certainly on paper 95% effectiveness of a human job is something that should make your eyes pop out, but at the same time, maybe you've been on a chatbot and you're like I want to cancel my subscription and it says, it responds with oh, I want to upgrade my subscription and you're like, no. And so maybe you go through that whole process and eventually you wind up talking to a human being to get this all resolved.

And I've had similar things happen with chatbots where that first amount of time was completely wasted by me. I wasted my time with the chatbot, and it gets me into the mindset of I want to just not even deal with chatbots, at least with this company, and I want to go straight to a person. That 5% accuracy difference can absolutely kill a project within artificial intelligence. And it's not that you're always going to get to a 100%, but you need to know how you're going to handle it when it doesn't. And so you really kind of shift with artificial intelligence from not just you build it, you build it well, and then your focus really becomes on exception management. So you have to focus your vision on handling cases where the AI hasn't worked quite well, you need to have those really optimized whereas right now it only accounts for 5% of your business, but then that 5% becomes a 100% of what you need to focus on. So I would say the biggest issue there would be switching the focus to managing those exceptions.

Yolanda Muchnik:
That's a great point. And as somebody who has engaged with many a chatbot, I can attest to the fact that it can really turn you off from a company. So, great point. Well, one of the broader narratives I've seen when it comes to AI touches on the ethical and legal boundaries of technology. And so, here I want to talk a little bit about with you how, if at all, do you see these kinds of debates playing out in multifamily? Are the use cases we're seeing in multifamily more or less prone to these kinds of concerns?

Chris Snyder:
I would say less prone, especially in comparison to what we have seen actually done in the world. So we've seen, for instance, deep fake videos weaponized. We've seen convincing people on social media that people have said things or did things that they did not do. We have seen all sorts of misinformation that has been generated by AI because AI can make it look very realistic. So I don't know that we're going to have some of those really foundational problems within real estate that we have in the world at large. But specific to AI and multifamily, I'll say that it is extremely important that we take a defensive mode when it comes to AI. And that doesn't really stop you from implementing AI, it just means that you have to think about the ramifications of how it could be used. You have to put on your white hat and pretend like you're a bad actor or ask yourself and have a whole discovery session around what could go wrong.

So I could think of a number of scenarios where this applies. I think one of the best ones that I could think of is smart homes, so smart multifamily complex. So hopefully, we can see some great benefits from AI and smart homes, especially when you have multiple families and you can do things like have AI control so that it knows when people are coming home so they can have the air conditioners, not all turned on at the same time, or maybe adjust for rates and monitor water usage and have security around maybe when a maintenance person can access a smart lock and get into an apartment complex and monitor that. There are tons of benefits. AI and smart home are coupled really well, but what you don't want to do is you don't want a bad actor or some sort of bug or anything like that, have the ability to, for instance, unlock all the doors in the complex at once and leave the gate wide open.

And so that's an obvious statement right there that you don't want that to happen, but you have to be very defensive in your approach for how you want to implement AI. You have to think what can go wrong and what would a person who has bad intentions use this tool? How would they use this tool in a malicious way or in a way that can cause serious damage? And so it has to be a part of the conversation at all points. Even if you're buying from a vendor, you need to ask that vendor those questions, how do you prevent this from happening? So I hope that answered your question.

Yolanda Muchnik:
Yeah, yeah and so well put, thanks for that. Well, Chris, we actually work together and I hear you talk a lot about how AI is changing your own team's approach to product development and engineering. And so can you let our audience in on that just a little bit and maybe also share where else you see potential applications maybe within Zego's own product roadmap?

Chris Snyder:
Sure. So a few things on that. So we just had a hackathon. So a hackathon is where we take a couple of days and all of the engineers that want to, this is opt in, they can join the hackathon. They set up teams. We had about 25 teams of engineers and we were very focused on two things, one is artificial intelligence, two unique items focused on artificial intelligence. And we specifically made this hackathon. We've had them in the past. We made this one focused on engineering. That way, they knew their own domain and they knew the tools and how to use them. It was not focused on product. And the purpose for this is we really want to get into a train-the-trainer kind of mentality because it really reverses the roadmap when something new like artificial intelligence comes into play. It really changes it from coming from a customer, whether it's in support or through marketing or through email campaigns or whatever. It really changes that from them providing us information to them asking us what are the capabilities?

So we figured we need to get the engineering teams up to speed on that question. We need to make sure that we have actually coded, applied for it, found any problems and can guide people on artificial intelligence. So it was a great hackathon. It was absolutely... we've been doing these for years and years, and I would say this is the best hackathon that we've had yet. But I'll give a couple of examples of what we did and it may spur some ideas and some questions from people, I really hope it does. Two of the winners were in AI categories. We had more winners, almost half of the entries were artificial intelligence oriented. One of them was fairly simple conceptually. We have a problem with documentation, and you might call more accurately a problem with documentation, a problem with knowledge.

Everybody wants knowledge. Nobody wants to make the same mistakes that everybody else did. Documenting mistakes when you're making them or things that you're learning as you're learning them can be a little bit difficult, especially in a way that is shared and keeping them up to date is near impossible. It's a common problem in engineering everywhere in the world. So one of the entries, which was brilliant, took our own private instance of artificial intelligence engine, and by private it means that our information was not shared with the rest of the world, it was not absorbed by the artificial intelligence engine for future use. In fact, it was all deleted and has completely disappeared in the aftermath. Anyway, he took all of our software code and a bunch of our product documentation, folders worth of documentation, from a product standpoint and developed a natural language engine. So you could ask it as just a general question, I'll tell you about one of our great features, one of the ones that I love 'cause this is where we tested it, is we have in Mobile Doorman we have something called content tiles.

Now these are tiles where you can swipe right to left. They're just little images and the first one could be like a coupon for 20% off at Domino's, the next one could be hey, we're having an event at the apartment complex pool on Tuesday and the next tile could be something like press here to pay your rent. So you could ask the engine that was built, what is a content tile? And it would describe it better than I just described it, using its own language, using descriptions that it had pulled from product documentations. And that's great. The chatbots can already do that. But you could also ask it technical questions. You could say you don't have to say that you're an engineer, but you could say what parameters does the mobile device pass to the content tile application at the backend that helps us to render these tiles? And it would give you the name of the parameters and the scope and all of the sort of technical detail.

So it had a comprehensive knowledge, it had the product knowledge and it had the engineering knowledge. You could add financial knowledge. How much money did it cost to build? How much has the product grown since then? All sorts of things. You can throw the sources at it and communicate with it in a natural language. That was one of the two winners.

The other one is the future of artificial intelligence. And this is why you hear so much about ChatGPT. The G in GPT stands for generative. So it means it's not just taking a question and returning knowledge. It's not just got a simple ML machine learning engine behind it to give you an answer, it's actually doing something. It's creating something. And so you've seen people will say, write me a product documentation, but make it sound like William Faulkner wrote it or draw me a picture of what it would look like if Arnold Schwarzenegger and a golf ball had a child. You can go crazy with all of these things, have it do all sorts of Photoshopping. The one that won the contest, it took detailed requirements for a product, for a new development and actually wrote the scaffolding for code. It could have gone further and actually written a lot of the code, but we put a limit on where it was close to a 100% accurate, and so it did.

This engine took maybe 10 minutes to do what would take, what you would normally assign to a junior programmer and it would probably take them about two weeks to do. So the generative, the creating of things is extremely valuable. And it can learn by looking at your previous code. It can learn by looking at your historical stuff, any sort of constraints or stuff you have in technical documents or in product documents. Generative is why AI has resurfaced. My dad did AI in the 1970s. AI has been around for a long time. It's been in movies since the '70s. I believe the 2001, A Space Odyssey came out in the '70s. That is the future of where it's going to be. It is going to be able to do people's work and allow them to focus instead of maintaining a business to growing a business and that's where you get some great economic benefits out of artificial intelligence.

Yolanda Muchnik:
That's awesome. And I look forward to continuing to keep tabs on how your team is using AI and maybe even how we can incorporate it and within our own product roadmap moving forward. So Chris, as we begin to close out this conversation, I want to end with some actionable advice for our audience. Do you have any tips for operators who are new to the concept of artificial intelligence and might be considering how it could be applied at their company and their properties?

Chris Snyder:
Sure. So I would say the biggest and most important thing is think differently. So there's so many quotes on the internet that are mis-attributed and never actually occurred. And so I'm going to do a couple of them and I'm going to tell you who they're attributed to, even though they probably never actually said it in real life. Henry Ford supposedly once said, "If I'd asked the people what they wanted, they would've said faster horses" and along the same lines, Steve Jobs supposedly said something along the lines with regards to the iPhone, "If I'd asked the users what they wanted, they would've asked for a better stylus." So for people to be successful with artificial intelligence, realize that it's a new world and so you have to, again, to borrow from Steve Jobs, you have to think different. You have to think differently. If you use AI to do human processes in a human-like way, you're not getting a lot of benefit out of artificial intelligence. You might reduce some headcount, but the fact is that AI is more powerful in some areas and less powerful in some areas.

And so actually, to be successful, you'll need to rethink business processes. You'll need to think how did we do this? Because if you think that an artificial intelligence device is just a substitute for a human, that's where you're not going to get the benefits from it. It does do things differently. It can do things that humans can't do, and so you need to take advantage of that. So I would say the biggest one is think differently, think defensively.

Yolanda Muchnik:
Awesome. Well, Chris, you are one of the smartest and funniest people I know and I am just so thankful that you took time to come join me and chat with me today. Thanks again for coming on the show, and I can't wait for our audience to hear this episode.

Chris Snyder:
I'm looking forward to it as well. If you'll just edit out any ums that I said, I'd be more than appreciative. We have an AI engine that'll do it for you.

Yolanda Muchnik:
I love it. Will do.

Chris Snyder:
Alright, thank you.