AI Isn’t the Problem. How You Manage It Is.

0:20All right, ladies and gentlemen of Good Morning Hospitality. We are back with a special guest who is not a stranger to the show. She has been on many times and we're excited to discuss AI and what does hybrid 2.0 look like in this world today. So, Varad, thank you so much for joining me. Always my pleasure, Will. Of course. Well, for the listeners, sorry to spoil it, but this is not a live episode. We're actually recording pre-guest eval, but this is being released after your main session on the stage in Madrid, which me and the whole GMH team is so excited to attend. And I'm just really excited about this concept that you have built around AI and how not only is it just a model problem, but it's actually a management problem.

1:11And it's a different take. I'm not going to lie. It's something different that we're hearing. And so I want to just give you the floor to kind of explain the concept of what you discussed on stage and then we'll get into some questions and kind of break it down for the listeners. Yeah. So again, we've had technology issues over and over again in the past couple of decades. And it's always about people and not just technology. It's about adoption. It's about change management. And I claim the same goes with AI even more so, and I'll explain why. And we talk about a hybrid team because, you know, starting COVID, we had the hybrid team of home, office, work from anywhere.

1:59How do you manage that change instead of having everyone together in the same office? All of a sudden, you're in different places. And to me, the hybrid 2.0 is people in AI. How do they work together? How do you redesign your team, your work, your processes, your methodologies, everything around the fact that your team is not just the people you know, it's people and agents or other concepts of AI that I can elaborate on. But it's not the same workforce and you need to redesign your work again for that new hybrid model. So that's what I mean by hybrid 2.0. Yeah, no, I love it. And I wanted to get into it because I think most operators as AI, you know, ChatGPT launched a couple of years ago.

2:52They just started buying all the AI tools, AI pricing, AI optimization for their listings in terms of description and maybe photo enhancements and all the things. But I think we're all starting to see like, where's the ROI? Like what actual AI tool is providing to my bottom line? not just making things faster, but actually being better and then what I was already producing myself. And so I kind of want to ask you about this, you know, management piece here, because a lot of operators as, you know, they have 10 employees, maybe 20 employees, let's say they have a couple hundred homes. Is this like management, not tooling piece of your discussion really for the bigger enterprise players or can the actual smaller operators who are scaling and building their business keep up in that term?

3:45To be honest, I think it's even more important for the small operators, and I'll explain why. Because if you have 100 people on your team, you have redundancies, you have other means to operate, you can also build on top of platforms like Guesty using VMCP. But when you're a small operator and you only have one marketing person, right, and one operations person and one finance person, all of a sudden, using AI, you can enhance their capabilities and allow them to work as if they were two or three people on the team. And that's much more impactful than, you know, reducing 10 people out of 100, right?

4:30And I'm not even talking about reduction. I'm actually talking about how do you create more value and allow your team members to extract much more out of technology, how to enable business growth with the same team? How do you become more effective, a better communicator? That's what we enable to do with the iCapabilities. And in Guesty, we have tried to build it so it's kind of an AI in a box with our agent hub. So you don't really have to start imagining and developing. You really can tailor the agents and solutions to the needs of your business. And it could be a small business. And then you'd probably adopt slightly different agents and tools than a big business.

5:19And that's absolutely fine. Yeah. Well, we all know that, at least for LLMs, right? like it's only as good the response that you're going to get is only as good as the prompt and the information that you feed it so i kind of want to push back in terms of like how you say the management is more important in the systems which i agree because i think every business knows that they whether they have one employee or they're a couple contractors to a hundred um how you operate in the systems you build for certain processes and and things that you do is really really has to be updated and kept and maintained and well managed.

5:54But I also think, you know, the models are pretty important. And I don't think the models are good enough. I don't think, I think I spend more time with Claude and ChatGBT correcting them sometimes than I'm actually getting actual effective work done. So where does the model actually come in play? Because I don't think it's there yet. And so I would love to know your guys' take. So I disagree. I think the the problem is in many cases of hallucinations or wrong answers is due to the fact that you don't give the full context and don't connect the right data sources to your LLM. So if you don't give them the right context, then yes, they may, you know, make up answers that are not correct.

6:38But if you do, and that's what we try to do with Guesty, right? It's really having all the data, all the sources, all the context, all your history, past interactions with customers, past reservations, etc. If all that is including in the data of your AI agent, then yes, it will provide the answer quickly, sufficiently, accurately, better than a human. So the question is, how do you implement? How do you really document the process? Right? How do you document the handovers human to AI and AI to human. And if you do that correctly, then yes, AI would be better than human. And in many cases, we do keep the human in the loop.

7:25And in that case, you have the best of both worlds. So if you think of the Guesty co-pilot, for example, you would have a human powered by AI in a sense. And if we think about our agent hub, then we really made those agents work sometimes alone, sometimes simultaneously with a human that will know to extract the right value from them, if that makes sense. Yeah, totally makes sense. I want to go into this concept because you guys call it the infinity team. And I love the phrasing. But before I ask you the question around that. I would love for you to just kind of maybe give an example of what that Infinity team looks like today, maybe with a guest user or how you guys imagine it to look as people start implementing agents throughout their business.

8:16Sure. So I think when I envisioned a hybrid team a year ago, when I thought about teams at Guesty or teams with our clients. I envisioned an org chart with a manager, and they have like four people and four agents. And that's not how it's playing out, right? What we understood is that agents may cover a certain task or a certain skill or a certain part of a role and not necessarily a full role of a full person. So when we designed the agent hub And when we think of AI implementation in general, what we do is really deconstruct the role, reimagine it and understand what parts of it should be human, what parts should be AI and how do they interact, right?

9:09Where is the handover? I think the co-pilot is a great example of that. We have a customer, LiveSuites, they have over 400 properties. And a nice thing is that they grew with us over 10x over their time with Guesty. And they think of the co-pilot as their super duper PA assistant, analyst, all that in one agent that really allows them not to do the thinking for them, but collect data for them, analyze it. I'll give you an example. If they want to know what properties in a certain region, for example, have the highest occupancy but low performing ADRs, right? They get that in an instant and can then feed it into their dynamic pricing and understand how they really create more value immediately by getting those analysis.

10:14So if they had to check it one by one, or, you know, just start looking at spreadsheets and dashboards, it would take them a ton of time. But you have your agent that is really your buddy that is really feeding you with the information, the data, the analysis, and allows you just to take that decision, but not spend the time on the research. Totally makes sense. I'm kind of curious, though, because I know a lot of people who are in these jobs, right? They're probably thinking about, okay, this is my task, right? I'm already doing this. But on your session, you talk about decomposing jobs and real-time judgment calls.

10:50And so I know it's all great in theory. I was a property manager once, but at 2 a.m. when I got the phone call about the pipes bursting or maybe some other catastrophe happening, one, I was a horrible property manager because I am the heaviest sleeper. So I never woke up for those calls. But how do you know when, you know, that judgment call at 2 a.m. has to happen? You know, these labels aren't aren't easy. They're not. Hospitality is not a easy, quick, like, all right, boom, Will's asleep. He's not waking up for the phone call. Let's have our agent step in. So I guess maybe walk me through that when, you know, this, you know, label from triage to judgment really needs to happen.

11:31So it starts with the property managers and their teams that really understand the work, really re-looking at the entire workflow and decomposing it into those skills or tasks and saying, I'll give you an example, answer a guest message, okay? It seems like one task, but it's not because you have to read the message, you have to get the context for the message. Do you have the information required to answer the guests? don't you? Is it a judgment call? Is it just a straight off answer? And you need to draft a response. You need to decide, is it going to be sent by AI or human? Does someone need to review it?

12:16So you have seven or 10 different tasks just in that one sentence of review a guest message, right? So what you need to do is really think of the different portions of it and understand what do you want to automate and whatnot. And that's what we've built with ReplyEye, for example, where you can decide on your level of certainty with allowing AI to respond on your behalf. So you can give certain questions 100 % AI response and certain questions say, no, this is something I need to review and respond, you know, click that last button and that's fine. But we allow you to really judge and make the decision on where you want the human to step in and where not.

13:07And you're right that in 90 % of the cases, AI can respond without a problem and it's going to be fine. But that 5 % or 10 % that are the human judgment, you need to know what exactly those are. How do you define them? How do you make sure you have a human in the loop at the right time in the right context. Yeah, I was going to say all the stuff like you were kind of talking about, like when you want the Wi-Fi or the lock code, that should be A, in the listing or like the check-in instructions that they get upon arrival or a couple hours before arrival, however that works. It should be in the PMS.

13:43Like all the information should be stored somewhere. So that's easy, 100 % AI response. But, you know, Varad, you and your partner on honeymoon or anniversary and the champagne bottle didn't get delivered. And let's just say maybe you brought your dog. If you have a dog and it got locked in the bathroom for some reason because the lock was not working. Yeah, like it should go to a human. I think it should have that human call because like, hey, maybe you're going to need maintenance or maybe you're going to need an emergency response. I don't know, like you're going to need something. And even to find the areas where a guest needs that 100 % trust, for example, finances or security, right?

14:27If someone says, I think someone's breaking into the apartment, right? They don't want an AI response. Someone needs to rush over there, right? So knowing to define what are those human moments is very important. 100%. 100%. Well, I also love kind of jumping around here. I kind of love what you guys say about AI owns the code and the team owns the hug. So I kind of, yeah, just break that down for me because I think it's, it's refreshing because I think that's what all operators want, right? Like we all, I've talked about this for years on my previous podcast and on this show that we all got into this space, not just to be, you know, cleaning toilets and flipping mattresses.

15:12Like we actually got in to provide hospitality, to create moments and experiences and make sure people have a good time, whether it's in the home or the destination. But yet, it seems like a lot of the time, especially through COVID, we were all just managing tech more than we were managing anything else. So yeah, kind of break down this concept. Yeah, I think it's especially interesting in the short-term rental industry. And AI has that particular advantage where we can really now focus on what matters on that human interaction and experience. Because the good thing about our industry is that it's here to stay, AI or not, right?

15:50People are going to want to travel, I hope, yeah, to real destinations with real people, even when AI can provide certain experiences. And we have the advantage to be in the center of that. So AI really allows you, and I mentioned a 5 or 10 % where human interaction really matters. It allows you to focus on that 5 to 10%. So really focus on the experience, on that hug the guest might need, on providing that local feeling when the guest comes to you in the morning and asks, hey, what do people do around here? What does this town specialize in? Having that real touch. But yeah, as I mentioned, if at 2 a.m.

16:37I want the door code or the Wi-Fi password, I don't need a hug. I don't need a person. I need an instant answer. I need it quickly. I need it now. And I'll be more satisfied with AI than a human at that time. And it's interesting when we checked our responses at Guesty, the combination of AI and human yielded a 5 % higher CSAT than just human. because, yeah, you know, when you need to, you get a faster response and when it's not relevant or when AI can't solve the issue, then it escalates to human. But in some cases, you don't need to get to that escalation. So that is the advantage. Yeah, I like that.

17:22Okay, you're kind of convincing me to be less skeptical about AI in our industry. But I will ask this last question to leave the listeners with a little bit of something to think about. from your talk on stage. And that's on the speed of adoption versus trust. And you kind of mentioned it a little bit earlier, but hospitality truly does run on trust, not only with the owner or the HOA, but with the guest. And I think even now, if we want to get a little bit deeper with the team, right? Like I want, as a team member, I want to be able to trust that my coworkers aren't going to just be offloading, you know, their work to Claude or ChatGPT and not really checking or verifying and then making my work even harder.

18:02So I think there's a big part of trust there. So what can we realistically expect without eroding trust in this speed adoption era of AI? Sure. And there are two parts of it, right? Trust vis-a-vis your owners or guests and trust vis-a-vis your team. And when you talk about your team, it's really about creating that curiosity and engagement and allowing them to understand AI, learn the capabilities, feel that you're empowering them. And then when there's more knowledge, there's more trust, right? And then, of course, there's how do you enable it and how do you help them grow? How do you help them adopt it?

18:54How do you help them really document those processes? And of course, test and pilot and redesign the way they work. And don't just do everything at once. Start with one skill, pilot it, see that it works, and then move on. So once you do it gradually and once you put people in that circle of trust of we're in this together, we're learning AI together, we're learning how to get better together, I think that creates a lot of trust. And of course, transparency is a part of it. Don't lie, not to them, not to the guest, not to the owner. If people ask you, is my work going to change? Am I in danger of losing my job?

19:40Don't say, don't worry about it. Explain. Say, some roles are going to change. We're going to change the way we work. It may impact your work, but it may make you do bigger and greater things and focus more time on valuable work rather than mundane work. but talk about it be very open about it with your team and if owners for example are concerned hey are you now going to do everything with AI I expect you to be in the loop then tell them yes I'm in the loop those are the KPIs that I look at this is how I'm getting better this is how I'm creating more value to you using AI and this is where I'm still in the picture and you can always trust me to be there for you and be there in the background, even if I implement AI tools.

20:33I like it. For Guesty, as the COO and president, I would love to know what implementations have you guys made internally for AI and just how you guys work and how you guys structure. I'm curious because you launched the agent hub, which I was really excited about because I was like, okay, it's the first agentic thing I've seen in hospitality so far, at least when you guys launched outside of like, you know, just Claude or, you know, ChatGPT. And so just kind of curious, what implementations have you guys made and how has that affected Guesty internally? Oh, wow. We've done so much. And I think the beauty of it is that we started over two years ago with AI adoption and really transforming and reimagining the way we work internally.

21:21And that enabled us to come with the right AI mindset to developing our product and to working with our customers. So, for example, our development is all AI based. And currently, the committed code that is solely AI driven is 85 % of the code. And of course, all our developers work with that. And we've changed our product development processes around that. We changed, of course, the customer experience around that. And we talked about human in the loop. Yes, we have automations and we have our very own agent responding to some standard customer questions. But we escalate to humans where necessary.

22:11And we have the super agent that is our internal co-pilot that works with our agents, our human agents, and helps them respond quickly. because I can get a very complicated question about financial or accounting issues. But with the super agent, I get all the data at my fingertips automatically in one minute. And I can respond as a human much faster and much more accurately to the customer asking the questions. So those are just a couple of examples. But we've really transformed every department, including the marketing department. and you'll see it in Guestival, but a lot of the work was made with AI as well.

22:54Well, not to give any of our listeners or viewers any FOMO, but getting to see everything prior to the event, obviously this is coming out during the event, so I'll probably be even more surprised on site and seeing it all in action, but I don't know how you guys do it. I've heard all in the past, Guestival is one of the best industry conferences to go to. It's not one of the best. It is the best. And I can say that after going to Mexico City last year, it was incredible. The hospitality, the experience you guys built around the destination, to be honest. It was very Mexico City. I felt like I was in one of the greatest destinations to experience the best local tacos and the best tequila for those who like tequila.

23:42And so seeing Madrid, I'm like, oh my gosh, this looks very Spain. So I'm very excited. But yeah, it's going to be good. So if you missed out on Guest Eval, hopefully you won't next year. So make sure to keep the FOMO going a little bit. But Barrett, thank you so much for taking the time to sit down with me today in this episode. I know our listeners are always curious about AI, agentic AI, how this is really going to impact their role, their businesses, their systems, their processes, and of course, the guest experience. So I think you gave some great insight and i love uh what you talked about on stage so for the listeners don't miss out on more to uh more to come thank you and i think an important takeaway is at the end of the day 70 is people even with ai implementation it's all about you your team how you manage them how you help them get into this new ai era so looking forward to some exciting times yes well we'll We'll have you back on the show as things continue to develop, obviously.

24:45But thank you for your time and for the listeners. Make sure to grab all the links in the show notes. Make sure that you don't miss out on Guest Uval next year. And of course, like always, we'll see you guys again next week.

25:12Mm-mm, mm-mm, mm-mm, oh