Taking Agentic AI Beyond Chatbots With EliseAI

[00:00:00] The leading issue of agentic AI in businesses right now is ensuring agents act with compliance guidelines. And Denodo applies guardrails across your entire data estate. By aligning your company's data infrastructure under one system, these guardrails perform consistently across your platform. So start scaling your business and start with Denodo. Simply visit denodo.com to learn more.

[00:00:32] What does agentic AI really look like once it exits the keynote stage left and starts dealing with rent payments, leasing inquiries, maintenance requests, and very real people who simply want an answer before they lose the will to live? Well, my guest today leads client strategy at a company called EliseAI. And the company builds AI agents for housing and healthcare.

[00:00:59] And with its technology now operating across over 6 million housing units in the United States and Canada, he's got quite a unique story to share. So he's going to bring with him his unusual perspective because he spent over a decade working in multifamily housing operations and was previously an EliseAI customer. But he has used this technology from the operator's side.

[00:01:27] And now he finds himself working with the teams deploying it. So today we're going to discuss what separates an AI agent from a chatbot, why completing an entire workflow requires context from several systems, and explore what happens when a rent collection conversation is collected to a unresolved maintenance request. And Jacob will also share a surprising lesson about customer acceptance of AI.

[00:01:55] And we'll learn how residents appear far less concerned about whether they're speaking with a person or AI, because all they really care about is whether their problem is resolved quickly and accurately. There's going to be a lot in this episode for leaders outside of housing too. Every business has repetitive work and complicated workflows. Every business has repetitive work and complicated workflows. And customers who would rather receive a useful answer than hear your call is important to us

[00:02:25] as they're waiting on hold for 40 minutes. But it doesn't have to be like that. And at that point, it's time for me to officially introduce you to my guest, where we'll talk about all this and much more. So thank you for joining me all the way from Chicago today. Can you tell everyone listening a little about who you are and what you do? Yeah, thanks for having me, Neil. My name is Jacob Kosher. I lead client strategy at Elise AI.

[00:02:52] We're a vertical AI company serving two of life's most critical areas, housing and healthcare. If we were going to deploy vertical AI in any big industry to disrupt it, what better spaces than housing and healthcare? But I've got a unique background on the Elise AI team. Before joining the team here, I spent over a decade in multifamily housing operations, where I was actually a client of Elise AI. So I approached the technology in our day-to-day work as somebody who's actually used the tech to transform what we were doing in the past.

[00:03:19] Highly manual, repetitive processes that today, with the power of technology powered by Elise AI, we get to automate with this remarkable new technology. And I don't know if you know the answer to this, but a question I've got to ask before we go any further is the name Elise AI. Where did the name come from? Yeah, all credit to our founders, Min and Tony, who founded us nine years ago and are still with us today. As you can imagine, we started in the multifamily housing space, a lot of leasing to be done.

[00:03:47] So once upon a time, we were known as Meet Elise. That's when we were just handling leasing inquiries. And as we've since grown to cover the entirety of the renter lifecycle, it's going beyond needing a lease. And now we're managing all aspects of the leasing process, but also residents on site and all their various needs. So kind of the name evolved as our platform evolved, but all with the goal of helping multifamily operating where leasing solves all problems.

[00:04:12] So if we can solve the leasing part of it with Elise, then we've opened up the doors to solve a lot more challenges for our clients. Love it. Why didn't I say that? It was staring me right in the face. Love it. But I mean, here, if we look outside, so much of the conversation around a Gentic AI that we're seeing everywhere at the moment is focusing on what it might eventually do. We seem to be in the prep zone at the moment, but you're already seeing agents operating across roughly one in six US apartments.

[00:04:43] So tell me, what does a Gentic AI actually look like when it's running very real workflows every single day at that kind of scale? What are you seeing? Yeah, well, you're right. I think we're unique in that we are already making a difference in the lives of our clients with a Gentic AI. We're now live at more than 6 million units across the United States and in Canada as well. But to be honest with you, Neil, it's a lot less flashy than what you'd expect or what you might read in the news.

[00:05:09] But I think it exemplifies the use cases that exist out there for AI. I used to always train my teams who were utilizing the technology on site to say, Gentic AI actually works really well for our workflows. Because in multifamily housing, it's all about repetitive work that's being done at scale. A 200-unit apartment community, as is very common here in the States, every month we've got to be following up with people to pay their rent. Every day where an inquiry comes in, we've got to be following up with that prospect about it.

[00:05:36] Every time there's a maintenance need on site, we've got to be following up with that individual about the status of their work order. So I say it's a lot less flashy, but it looks like all these conversations that we're having day in and day out with our prospective renters, with our current residents on site, repetitive, whether that's the beginning of the month or somebody has to renew their lease or they've got a maintenance work, it's an AI agent that's handling that at scale repetitively. So we in the multifamily housing industry can always focus on what we tout, which is the experience.

[00:06:05] That for too long now, we've been having to say, we've got to sacrifice the experience to deal with the workload. And now we've actually got a solution for that workload, thanks to all of the AI agents that Elise AI is powering. Yeah. And I think when it comes to AI and AI agents, it almost divides everyone straight away. There'll be the early adopters, they're just convinced that it can do absolutely everything. And then there'll be those that think it's, oh, it's just a glorified chat, but want nothing to do with it.

[00:06:32] Then you've got the ones that want to experiment, but as soon as they're asked to hand over the keys to their, I don't know, their inbox, their files, their calendar, then they get a little bit nervous and back off. And you could argue the reality is somewhere much more interesting, but what does it mean for an agent to complete an entire workflow from start to finish? Because we've seen all the bad stuff out there. It'd be great to celebrate, to focus in on some of the good stuff that people are not seeing or hearing much about. Yeah.

[00:06:59] Well, I think for us, that means the AI is able to handle that customer from start to finish without a team member having to step in. And we're very conscious of the technology, especially in the housing space and also in healthcare and all the rules around that, that when the AI is not able to answer a question or it gets to a point that it stopped, we need to have that handoff process there as well. But our goal we're building towards is being able to handle that entire workflow from start to finish, whether that's a renter who's inquiring about an available unit or

[00:07:28] someone whose lease is coming up for renewal and we've got to go through that renewal process. We see it as the AI is going to have a great integration to begin with. It's got to have the context of its work in order to be able to handle that workflow from start to finish. But it's also got to have context on other workflows. It's very common that we'll have somebody who's interacting with, say, a delinquency AI agent that's there to collect rent, but they have a work order as well.

[00:07:56] They're holding back their rent possibly because they've got a maintenance need within their unit. So in that instance, and this is what we look to to say, we need one AI that's going to handle both that rent collection conversation and that maintenance need simultaneously so that we complete the entirety of the workflow. And that only comes with a lot of time and a lot of experience in the state, in the market, where we're lucky enough to have been here, been in operations for nine years now,

[00:08:21] much longer than others in the space and have over a billion conversations that we built upon to understand all of those edge cases that exist so that our AI is no longer encountering things for the first time. So it can now handle that from start to finish with all the nuance and context that's required to make it a great experience for the customer. And I suspect everyone listening, no matter what business or industry, and they will think that their workflow is unique and complex, not like others.

[00:08:50] But of course, every workflow is ultimately the same. And property management is a complex workflow in its own. I mean, it involves leasing inquiries, scheduling hours, renewals, maintenance, and countless other everyday interactions. So which workflows have you found proved best suited to AI agents? And where do you find that humans still need to be firmly involved? Yeah, well, I think you hit a lot of them there. It's leasing inquiries, it's scheduling of tours, it's renewal conversations, maintenance triage.

[00:09:20] It's ones that I point to a lot. Leasing is a great example. When I was on the operations side, we used to train our teams that for every prospect or every interested renter, we needed to follow up with them five or six times to keep them engaged. Now the data shows that's skewing towards eight or nine times because we've got these little computers in our pockets that are built to keep us distracted. So what used to be an already manual process following up with a customer five times has now become even more work. Now we're having to follow up with them eight or nine times.

[00:09:48] So the workload has gotten greater and greater, but a great example of where we can utilize agentic AI. So leasing inquiries as you hit on there. Another one that I always point to, maintenance requests. In the States here, we launched resident portals to say, you've got a maintenance need, we'll log into your portal and tell us about it to submit a ticket. But what that meant for operators was that the data that we were basing all of our maintenance

[00:10:16] decisions on was only as good as what the renters were providing, what an individual renter was going into their portal and submitting. Now with AI there, we can not only service that customer more comprehensively, offer up self-help for them or do some more discovery, ask them for media files as part of that. But now we've got much more data clarity as well. We're not relying on the resident providing information. We've got agentic AI there to go and collect some of that information to give us more data on which to make decisions.

[00:10:44] So I like that example because it shows how we can both better serve the customer, that resident who's living with us, who just has a maintenance need, while also making sure that we've got better data on which to make our business decisions and hopefully be more proactive in our maintenance planning. And a couple of years ago, there were a lot of stories of how the majority of enterprise AI projects were failing to deliver on their expected results. I think there was a hugely exaggerated claim that 95% were failing.

[00:11:13] I think it was an MIT study. I mean, it got people's attention. I don't know how accurate that was. But I mean, from the deployments that you're seeing firsthand, what is separating an AI agent that becomes part of everyday operations for ones that never get beyond experimentation and get stuck in pilot purgatory? Yeah, those ones that aren't succeeding now are ones that are just kind of a chatbot feature that's being layered on top of the tech stack. So it's really just serving there as kind of a chatbot to answer questions.

[00:11:41] But unless it's kind of baked into the technology stack so that it has the integrations necessary and the context necessary to help that customer, then it's not really going beyond a chatbot. And we do a lot of education with our clients to talk to them about how the technology is evolving, that we've moved from kind of generative A or conversational AI to now agentic AI. And what does that mean for us? Because I found time and again that if we can be educating our clients on the technology,

[00:12:10] no one when I was in the space, no one was teaching me about the technology to begin with. But if we can educate the industry on how the technology is evolving and how it is moving beyond this chatbot that sits on top of your tech stack to one that's actually integrated with your technology and using all of the data and context and systems to support customers or refine our data, as I mentioned on the maintenance side, that's what's been the differentiator between one that just kind of serves its limited purpose there as a chatbot and one that actually

[00:12:39] allows you to rethink who's doing what work where and where can you best utilize your most important resource, which is your human capital. And before you join me today, I was doing a little research on you guys. And one of your more surprising findings that really stood out to me is that most residents don't particularly care whether they're dealing with a human or AI, providing, of course, that they receive a fast and accurate answer. That's probably the most important part.

[00:13:07] But what does that tell us about what customers actually value? And our business is sometimes maybe asking the wrong questions about AI acceptance. Yeah, I always used to say whenever I was having a bad day on the operation side, I would just go read some delinquency AI conversations. Because you can tell from the residents who are interacting with it that they don't always know it's AI. Because if they did, they wouldn't be communicating with it the way that they were having actual full-on conversations with it. But you're right. What that's shown us is that at the end of the day, our customers just want their problems resolved.

[00:13:37] And if that means they can do it in a matter of minutes or seconds with AI versus waiting two hours for a follow-up from an on-site team member, well, then they'll opt for the quicker solution. Because our residents, our customers, they're as busy as we all are. And they just want a fast resolution to that. And I'm glad you saw that stat. What really blew my eyes with it was when we started to launch different voices for our voice AI product as we started to introduce some regional voices.

[00:14:03] So here in the States, you could have a southern voice or a midwestern voice or a coastal voice. And those actually really moved the needle on the engagement with the AI. More people were going back and forth and having a conversation. But it was also leading to higher conversions. And so what that told us is that they were okay talking to AI even on the phone. It was just a matter of hearing something that was familiar to them. You know, the AI, as much as it's in the headlines right now, it is still foreign to most people.

[00:14:31] But if we can make even small tweaks to the product like that so it sounds less foreign, it's more common to them, well, great. Then we can service those customers even better. But time and time again, what the data has shown us is that customers just want to be helped on the channel with which they want to communicate with us as quickly as possible. And that's our goal, as we talked about earlier, to automate that entire workflow so the AI can handle it. And hopefully we can leave that customer satisfied and resolved and moved on to the next one that's queued up for us.

[00:15:00] And you've been on quite a journey here. And I suspect you don't always look back at how far you've come. But you've seen conversations with operators change incredibly rapidly from should we use AI to, hey, how do we get more value from AI that we already have? And then from there, what happens after the first successful deployment? And how should companies or people listening to decide which workflows to automate next? Because I think once you tick a couple of boxes and you realize the power, it kind of unlocks that light bulb moment.

[00:15:30] Where do you go from there? Yeah. Well, I'd say you're absolutely right. We see all the time. And one of my favorite products that we offer is delinquency AI. I got the opportunity to co-develop that with Elise AI when I was on the operator side. But I love that product because we have team members on site collecting rent. But what I found time and time again, whether our teams were willing to tell us or not, was that that was just an uncomfortable conversation for a lot of them.

[00:15:55] Calling a resident and telling them, you owe us money, that's something that's uncomfortable for a lot of people. So when we can deploy that, it's really easy to see the results. One team member on site, they can only call one resident at a time. With Voice AI, we can call the entire building at once to do it. So operators start to see, okay, there's a natural lift in performance because we've overcome this kind of human limitation or things that we don't naturally like to do. But we can also do it at scale like never before with mass communications the entire building.

[00:16:24] So that starts to unlock for operators like, okay, this is solving more problems than I even thought exists. I didn't think about my team members on site being uncomfortable doing this work or being limited in the scale that they can do. So once they start to see it in action, it starts to click. Then as you mentioned, the next step is, okay, well, what's next? What else can we tackle? And how do we start to transform our business from there? And that's what we see happening in the multifamily housing space right now, at least with American

[00:16:49] operations, very different than how operations work in the UK or other markets where we're starting to centralize to say, do we just have less work to do overall? And now maybe we can move some of that work to different, more specialized teams. We see that as the next evolution with AI tools to not just, great, we're starting to handle some of that work. But now what we're really working towards is unlocking the opportunity for you to transform your business and rethink the org chart or what work gets done where as we're able to

[00:17:19] automate more and more of those processes. And when AI agents start communicating with customers and taking actions on behalf of business, reliability and accountability become so much more important. And we will have a few people worried about, I don't know, an agent going rogue on a Sunday afternoon and saying things he shouldn't. So what have you learned about the guardrails, escalation paths, and human oversight needed to operate agents safely at such a scale? Yeah, we need to have that off-ramp when it exists.

[00:17:49] When the AI runs into something that it either doesn't know the answer to or isn't well-suited to handle, something like a resident who says, hey, I've got this charge on my ledger. What is this charge? I'm not going to pay this charge. We're not expecting the AI in its current state today to be able to handle that resolution. So we need to have that off-ramp. And we see that as, all right, that's a handoff to a team member. It's a break in that process. We want to automate the entirety of it, but we need that off-ramp.

[00:18:15] And so that's what we're working towards now to say, great, when we're handing that off to someone, who is best suited to help that customer at this time? Because as I mentioned, as we're rethinking the business or rethinking the org structure and who does what work, where increasingly we're now giving operators the ability to say, when the AI runs into this issue and it needs to hand it off to a human, which human does it need to be handed off to? Because now we can be very specific on who gets that work. And that's all with the goal of being more efficient.

[00:18:44] But that's something that we'd point to, to say, well, that's our guardrail. You're human capital. You're talented team members. Those are your off-ramps when the AI needs some support. But let's not just hand it to the entirety team. Let's hand it to Neil, who's best suited to help this customer. And we're going to hand the other one to Jacob over here. So we're really excited that we've got a good process there as part of it, while also allowing operators to rethink, again, who's doing that work, but who's best suited to help this customer, given where they're at and what exact need they have.

[00:19:12] So for any business leader that could be listening today outside of the housing industry, what lessons about deploying AI agents across millions of real-world customer interactions can maybe they take away and apply when deciding where Agentec AI could genuinely improve their own operations? Because I think it is industry agnostic that everybody's got a workflow. But any advice there based on your experience?

[00:19:38] Yeah, I'd say it's partly it's not just that the AI can have a good conversation. It's got to solve a problem in a really satisfactory way for the customer. So we make sure that, say, when we're demoing the products of our clients, we're stress testing it. We're not giving it an easy question to ask because we can have a good conversation with it. But big differentiator number one is, yes, it can sound smart, but it also needs to solve that problem for the customer. So how can we throw it some curveballs to make sure that it's doing it?

[00:20:05] The second I'd say is, and this has been our approach on the product suite, is we need to go really deep. And we need to make sure that we're solving one workflow and getting it as perfect as we can. We need to solve for all of those edge cases. So leasing inquiries that are coming in, well, hey, we need to solve when that's a broker who's looking for us. Or when that's a corporate rental provider that's looking for it. Or somebody who's looking further out than our typical customer. So yes, it needs to have a good conversation. Yes, it needs to help that customer. But it also needs to have the experience and the depth of the product to make sure that

[00:20:35] it's accounted for all the various edge cases that are inevitably going to come up along the way, especially if you're going to deploy it and try to make it as comprehensive as possible to really transform your business. And I'm not sure how much you can share here, but what's next for you guys? Where do you go from here? Or what excites you and makes you want to jump out of bed in the morning? Where you take this next? Yeah, well, I'm a bit of a nerd. So I do get excited about a big focus area for us right now, which is affordable housing.

[00:21:01] With all of the headlines right now, affordability more broadly, I've had the opportunity to work in affordable housing here in the States. And it still remains a highly manual process. Even things like renewing a lease agreement means that, all right, we've got to collect all your income documentation once again to make sure that you qualify to live with us for another year. So a big focus for us right now is understanding all of the workflows and solving for all of the edge cases in the affordable housing space. Because that's our North Star.

[00:21:28] How can we make housing more affordable and more effortless for all? And if we can tackle a highly manual process like the affordable housing space today, we think that sets us up to automate even more workflows in the future. So that's something that's personally of interest to mine, but also important to us as an organization as we're saying, yeah, we need to do these kind of start to finish workflows. But hey, let's take some of the most challenging in the affordable space, because if we can solve that, that makes everything else a little bit easier.

[00:21:55] Well, thank you so much for sitting down today and talking about how your technology handles the day to day for property managers, including answering leasing calls, scheduling tools, processing renewals, coordinating maintenance. And for anybody listening, no matter what industry and what is your workflow like, what could it improve on those cumbersome workflows that you think is unique and complex to your business? And for those people listening, if they want to get in touch with you or find out more

[00:22:22] information about Elise AI, where can they find you? And do you have a podcast? I've got to ask that as well, because you look great. You sound great. You've got the perfect backdrop and everything there. So is there a podcast that they should be listening to as well? There is a podcast. Yes, we have our In Good Company podcast. We're talking with multifamily operators about all these challenges you're hearing right now as we've got tech stacks that are becoming more complicated or operational org structures

[00:22:50] that are changing or even the change management required to get team members on board with this new technology and help them see it as a supplement to their team. All those conversations are more on our podcast, In Good Company, which you can find at Elise AI.com or you can find me on LinkedIn. Jacob Kosher can't help but posting about all the great conversations we're having on the podcast on my LinkedIn profile. Well, we see so much noise on our LinkedIn profiles, but hearing real stories and real value being generated and real success stories.

[00:23:19] I know this is going to be so valuable to people listening. So I urge them to check out the blog post associated with this episode over at techtalksnetwork.com. I'll include links to your podcast, your LinkedIn website and anything else I can find there that I think will be of interest. But thank you for sharing your story today. Really appreciate your time. Thanks, Neil. Appreciate the time. Good chatting with you. I love how Jacob removed so much of the mystery surrounding agentic AI there.

[00:23:44] Because at scale, AI agents do not always look like something from a sci-fi movie. Sometimes they simply schedule an apartment tour, collect the information needed for a maintenance request or follow up with someone whose lease needs renewing. And these tasks sound incredibly ordinary. And that is kind of the point. Ordinary work repeated millions of times.

[00:24:11] That is where automation produces meaningful results. And Jacob's advice also provides a useful test for every business. An AI agent should solve a real problem. It should connect with the systems containing the necessary context. Account for unusual cases, of course, and know when to transfer that conversation to a person. But a pleasant conversation alone is insufficient.

[00:24:38] Because a pleasant conversation alone is inefficient if the company still needs to call somebody afterwards. The point about customer expectations, I think, is something that I'll be taking away from this. People are busy. I am. You are. And we just want a fast, accurate resolution delivered in a way that feels somewhat familiar. I don't think there is a resident that's ever woken up hoping to spend the afternoon chasing an update about a leaking sink, pushing five for this, six for that.

[00:25:07] So a massive thank you to Jacob. And please check out Elise AI. You can find my guest on LinkedIn. Further conversations through the In Good Company podcast over at EliseAI.com. And over to you, a little bit of homework. Which repetitive customer workflow could AI complete today? And at what point would a person need to step in? Open up the notes on your phone. Have a think about that. Start writing down things.

[00:25:35] And let me know how you get on on your journey with AI in the workplace. But that is it for today. I'll be back again real soon with another guest. Massive thank you to Jacob. An even bigger thank you to you, not only for listening, but you've even reached the end. Achievement unlocked. Hopefully I'll speak with you all again real soon. Bye for now. Bye.