Episode 28 - Nerdy Insights with EJF - AI Edition

The information being provided is not intended to be legal advice. Legal advice must be tailored to the specific facts and circumstances of each individual case and an association specific governing documents. The information being provided is intended for general educational and entertainment purposes and should not be used to replace individualized advice of legal counsel. This podcast is not legal advice. Thiago. Hey, Dan. How are you? I'm here physically. How about you? Yeah, I'm here. I'm here. I'm ready to go. That's good. That's good. That's good. Another podcast. Another day. Another podcast. Yep. It's that. It's um. Live in the dream, man. This is great. Yeah. We got we have a great great show for everybody today. Yeah. We were doing. We were talking about things. You know, as we were getting tuned up for today and like talking about like AI, it just is exciting. It's scary, but is it scary? Yeah. No. That's scary. We have to wait. We're ready. Thiago. Well, maybe this will convince me to be a little bit more supportive. That the computer is aren't going to take over. You have to please it. Please. And thank you when you're doing your AI bots. We'll do that. Take over. Take over the world. You're in there. Good graces. Yep. So we have a we have two lovely guests today from EJF. Right, Thiago? Yep. We have two EJF real estate management. Excited to have them. We have Aaron Mendel, their COO and Mira Brown. That it's director of internal operations. We're welcome Aaron and Mira to the podcast. Thanks. Thanks guys. Excited here. Welcome. You guys know about AI? Just a bit. You know anything about generic to generative to a generic generative. Wait, wait, wait, wait, hold on. What is AI to begin with? AI. Alina and Alan. Alan. Time. We can we can go. Talk about practice. Talking about practice. Practice. Practice. What is AI? Aaron. Go. Sure. For those outside of Philadelphia who are Alan and I'm versus fans or not. AI is really, it's a large language model, right? So what they're doing is they're looking at all this data out on the web and saying based on everything that's been said, done and constructed over all these years. What's a good answer to the question you're posing to me, right? And that's in a short and sweet, right? Mira, I mean, I don't want to oversimplify it, but they're looking at. And I'm looking at the data. somebody. So I think that's going to be the big changes that we're finding. But with these processed driven systems like a gentick AI, you're still going to be a person, as Nero mentioned, put a guardrails up, right? It's actually going to be a little more involved than where we are now. I ask you a question, and I know from an illegal perspective, you guys say, write this brief for me, right? And it's site sources. And those sources don't exist. You've still got to do that due diligence to make sure that all those citations are correct. In this case, when we're building these processes, we've got to make sure that it doesn't go into something totally outlandish. And that's one of the things that Mira spent a lot of time on this year with our budget process, because that was something that we ran through this gentick AI system effectively where it created the budgets for our communities. A lot of tweaks to that and some learning curve issues, but in turn, back over to her again, the specifics, but that's what's not going to be replaced. The human needs to be there to say, this is what this looks like at the end of the day. And I got to help you figure out how to get there, because the system's not going to figure it out just yet by itself. And I think, Tia, I go, you said, you know, if I could get a bot to write my violation letters for me, yeah, the bot's going to write my letters for me now. And that's going to take an hour and a half or two hours out of my day. And now I'm just reviewing those letters. But what it lets me do is be a person and be an empathetic person to be on the phone with somebody who wants to talk about the violation letter that they got, for example, or to, you know, for me, I found a lot of what AI does for me and my role at HF is allows me to put the mundane to the side and do more future planning and more long-term thinking that, you know, a bot's never going to be able to do that. Or maybe it is. Maybe I just said to somebody on another call, don't ask the agent to do it. If the agent can't do it today, they might be able to do it next week, but like, he's pushing the limit of what you're asking it to do. So, you know, and yeah, Aaron said, we ran our budgets through our agent AI this year. And it was not perfect. And frankly, I spent a lot of hours on it, but by spending a lot of hours on it on our end, our managers got processes that were, you know, eight to ten hours before where they started last year, right? A whole business day of work before where they were last year, projecting out based on either, you know, I did some research on what Washington Gas said their increase was going to be this year and what DC water said their increase was going to be. And we pre-programmed that. So the managers don't have to go and look for that themselves. The other thing it did for us is it kind of unified EJF's output, right? There is a bottom line that we know exactly what was fed to our managers, because I have a four page five page thing that says, make this assumption, make this assumption, don't make this assumption that we gave to the agent. So you can train your agent, right? And tell it what you want it to do or say. There's something called a master prompt that we've just started looking at that you can, you tell it, this is what I want all responses from my organization to look like. For example, I want you to know that I want this kind of empathy or I don't want you to say, I hope this email finds you well, because that's how I know you used an agent to write it, right? But I hope this email does find you well, Mira. I assume you always do, Dan. So you don't have to tell me that. Yeah, fair. I love the idea that that like saving saving time on kind of wrote mundane tasks in order to free up time to do the other human things like be on the phone or whatever. And I guess a question I have is like, where do you draw the line between, okay, this is a good task for a robot versus this has to be done by a human. Because eventually, like you said, like we don't know what the capability is going to be in 5, 10, 15, 20 years. And what what this technology is going to be able to do. But like how do you, when you're looking at it from the management perspective, figure out these are good things, time saving things like that budget thing you just said about how you looked at watching and gas and you're like, okay, it's programmed into our AI for our budget for all our managers. Each individual manager doesn't have to go figure it out because I already figured it out and told the computer, which is I think what I understood how you explained it. I'm not exactly a big tech guy. So this is very helpful conversation for me, but how do you, how do you sort through, you know, where to start? What what the computer robots can be doing and what humans don't need to do? So I think it they are these repetitive processes that look the same across your portfolio that you can say, you know, everybody that's on Washington gas is getting a 3.4% increase this year. Or every, every invoice that comes from Pepco is for electricity. So those things that you can, as I said, set it and forget it for these things, the non-critical thinking things. Now I will say there are times when I go to my AI agent and I ask it some critical thinking things and see what it says, but that's where you're more likely to trip up. And so I think the other thing for me that's super important is it's garbage and garbage out. So if you have to make sure that you're putting in the right information into the agent for it to draw from. So our agent draws directly out of our software. And so I remember I had a conversation with the manager last week who was like, the agent told me this and it was wrong. And I said, well, did you ask the agent where it got it from? Because that's something that a lot of people who don't play in AI very often don't know. You can ask the AI, where did you get that? Like any agent, like chat GPT Gemini, but also our industry specific one that we use. You can say, site your sources and it'll tell you. So it's a good way to get ahead of what they call hallucinations when it is wrong. But when I went back to this manager who said the AI was wrong, I said, well, here's where it is in our software where that answer came from. So what we input into the software was wrong, not the agent. And I think that's something that's important to know is that the agent is only pulling from the information it can see. You know, a Gemini, a chat GPT is pulling from the whole internet. Both of you seem to have your hands out. Yeah, but I just want to say Halus, AI is hallucinating already. You guys, it's coming. Yeah. What? Oh, yeah. It was really funny. Well, here's what I was thinking about is that that little golden nugget mirror dropped a few minutes ago about how chat GPT only goes back to 2017. And then the concept of garbage in garbage out, I might submit to you that everything that's been put on the internet since 2017 is garbage. So then everything that's coming out of chat GPT is garbage. So I don't know. That's that's wild though. I'm still stuck on that from five minutes ago, 2017. I didn't know that, Mira. During the discussion on the capabilities of chat GPT, a claim was made regarding its knowledge cutoff date. Ms. Brown stated that chat GPT's training data only extended to 2017. This is incorrect. The generally accepted knowledge cutoff for the widely discussed base models is approximately September 2021 to early 2022. The original figure has been placed in the inaccurate information bin. We apologize for the error, but we do not apologize for the confidence with which it was delivered. Thank you. There's only so much data. And but it is everything forward. Like it's not but they had to draw the wine somewhere from where they built their first large language. Yeah, otherwise these data centers out in Loudon County, we exploding every day. I mean, I don't understand how this works. You guys, this is way over my head. That AWS outage was really fun to explain to a lot of people about what I mean, when you're talking about exploding data centers, how I had someone say to me, how does AWS's problem affect our payroll software? Right. And we had a learning opportunity there. So it's wild. So I think Aaron had mentioned earlier that this might slow hiring. It's going to then you know, we're mirrors explaining how this is going to be really great to kind of straight straight line or like, you know, make a lot easier. A lot of the tasks that are redundant in our day-to-day work as a community management team. But how does this going to like impact, you know, the management industry broadly? Like, you know, different management companies are going to develop your own software or create or you know, latch onto one of the other, you know, to another company that creates something very specific for for the work. And there's going to mean, there's obviously going to be impacts. I would expect it would help reduce costs that maybe we'll be able to, you know, get a bigger bang for our bucks. So we'll be able to start reducing the management fee. But we all know that once you start, once that starts going, it's like, you know, where's the end come? And I mean, I'm I'm here. I hear more often than I would expect otherwise that there already is a lot of price cutting and a little bit of, you know, debate in the industry world about what is what should fair pricing be in our management companies that are working in the areas, you know, throughout the country, kind of fighting like kind of creating a difficult market for innovation and like, providing good service. So I'm just curious, like, is AI going to push that train down, you know, towards like a broken, like something that's going to just like break down and make it worse. Are we going to be better? Is it somewhere in between? Like, how is this all going to shake out if you have to put your forecasting? We want to race to the bottom, Thiago. I don't think so. I mean, people talk about that, right? But I think at some point, you're going to hit a wall of folks who want to provide a good quality product. Like we do, right? Like everybody on this podcast do, right? We want to provide that to you. And you've got to hold the line and say, there's a cost to provide those services, because it's not all about the technology. Technology is a tool, right? And we have to remember that. And how we use it ethically, as well as from a financial perspective, is important, right? Because, you know, we have to make sure we can have an ongoing business. And it's going to help us with our costs. But it's going to free up a time to do other things, so that the relationship building doesn't go away. Because some people, while they might want to have a relationship with those computers, most don't, right? They need that human touch, and that's still there. And there's a cost for people, and that's not going to go down, because there's a value of my time, your time, everybody's right, that we have to set a line there. So I think that's going to stop the race to the bottom at some point. There's always going to be somebody who's going to want to take advantage of that and say, I'll undercut you, I just want the business. They're not going to survive very long. We know that, right? If you're not making it enough to cover your daily nut, you go out of business. That's not what we're looking for. We're looking to buy a good quality service that everybody's happy with, and use the technology to help supplement that and make it easier, because as times change, the demands are going to change, right? But in the early 80s, we weren't using this technology, and we were still using pages, no cell phones, right? I mean, it goes back a ways, but those of us who remember those good old days, meant we had time. You had to go find a pay phone, whatever the heck that is, right? You call somebody when you weren't home, and you're a real phone that you owned to say, what's the problem at your property? And they said, oh, there's a leak. Said, great, let me call the plumber. It's going to take a little while to get them out there, as opposed to now where everything's just in time. So I think AI is going to assist with that and make that easier and manage, because the system is being in place where it will be able to facilitate that just in time activity. But the end of the day, it's a tool that needs to be used in the grander scheme of what is property, community, asset management that we're all involved with these days. I think we should, there should be a day every year where everyone goes back to like either corded phones or wearing pages and beepers, just so we can remember. Or for some folks that have never experienced it, so they can actually understand what a phone book used to be. I have two members of our current staff's page or numbers still saving my cell phone. I accidentally texted to someone's page or number the other day. So, that's great. Well, I hope it sounds like maybe this, all this AI stuff, we'll get back to the good old days where you're cruising around and you don't have somebody calling your cell phone or blowing up your email constantly, because AI is taking care of a bunch of it. And maybe we'll have some more time on our hands to enjoy ourselves. That sounds good. Yeah, and I think, you know, with Aaron saying it's a tool, it kind of hit in my head, you can pay $5 to get, you know, an old band saw that you move your arm back and forth with. Or you pay $125 and get a circular saw that gets it done in half the time. But you have to pay more for the better tool, right? So in this race for the bottom discussion, everybody's monetizing AI, right? And so there's going to be a price for that as well. And so the AI plus the hundreds of years of experience and are executive leadership team, for example, you've got to pay for some of that. There's then, as Aaron said, the cost of people is not going down and this is a people industry, right? You're dealing with people. Most people is largest investment. You know, I think we're all very emotional about our homes. And so, and keeping that in mind, even if I can, I can give the agent instructions to keep that in mind when they're dealing with you. That's totally a possibility. But it doesn't change me getting on the phone and going, Aaron, I hear you. I understand there's a leak and we're going to make this right for you, right? That's still going to be part of our industry forever. I don't know that I have much more to sum up. That's, I have hope now, Dan, at the beginning, I'm like, I'm not sure. I'm not sure, but I'm feeling much better. Yeah. I'm feeling pretty good about it. Aaron and Mira really have swaged my concerns. Yeah. No, I'm feeling good. I appreciate that. Thanks, guys. That was very helpful and informative. Absolutely. I'm going to throw down the gauntlet to you too. I've done this with other people on our team. We'll spend five minutes a day interacting with an AI, pick chat, GPT, Gemini, or whatever, and start asking a question. Just I'll tell you one of my favorite uses I like to cook. I will say like, I have a bunch of basil in my garden right now. What should I make? Like, but those kinds of basic interactions and start seeing, try it out. Try it five minutes a day and you'll find it took me maybe a week before it's the most used app on my phone now. All right. I like that. I mean, I will, I will take up the challenge. You know, I think it's, it's a good idea getting used, getting used to using these tools because I am apprehensive. I am, I am a good, I, you know, but, but if I use it, I will probably like it and I'll get used to it and, and maybe that's a good thing. You know, Dan, you say you're a lullid, but you also like make your own gifts and send them out. I learned how to do a selfie gift recently. Did you guys know you can do a selfie gift on your phone, on your smartphone? I'm terrified of it because it does not turn out well of me, but yes, I've, I've had them said back to me. No, guys, please, please don't. Oh, my goodness. All right. Well, that, that was fun, guys. I, I think we're, I think we're probably all set unless anybody has any, um, parting words. No. No. Don't be scared of AI. That's my parting word. Mine's, mine's human in the loop. Yeah. Love it. Well, Miran, Aaron, thanks so much for joining us. We really appreciate the time and then of course all the great work. EGF does for the communities in the Northern Virginia district and beyond these days. Um, and so we appreciate again, all the services you all perform and for joining us for a little bit on this pod. Yeah. Thanks for having us. Wonderful time. Appreciate being here. You got it later nerds. Later. The information being provided is intended for general educational and entertainment purposes and should not be used to replace the individualized advice of legal counsel. This podcast is not legal advice.