How VAI Resort Built 12 Custom AI Apps
The assumption that the bottleneck for AI in hospitality is the tech. I think the bottleneck is really around the data and the judgment. AI on broken data is dangerous.
Speaker 2:From Hotel Tech Report, it's Hotel Tech Insider, a show about the future of hotels and the technology that powers them.
Speaker 3:Today, we're speaking with Bill Legrand, the executive director of commercial applications at Vai Resort in Glendale, Arizona. We wanted to bring Bill back on the podcast because he's had quite the year of experimentation and building with AI. Bill shares some great examples of how AI can save time and money. If you're curious about how AI can work in practice at your hotel, this episode is for you. Let's dive in.
Speaker 3:Well, thank you so much for joining us, Bill. Joining us again, we're thrilled to have you back on the podcast for a bit of an update and some interesting AI topics as we'll get into. For anyone who didn't hear your first episode, I would love for you to kick things off with an intro on you. Tell us about your role and your property, and we'll go from there.
Speaker 1:All right. Well, thanks, Adrian. Thanks for having me again. I'm Bill Legrand. I am the executive director of commercial applications at Vai Resort in Glendale, Arizona.
Speaker 1:We are a pre opening integrated resort with some cool features. We've got 11,000 rooms spread across four towers. But integrated with one of the towers is a 14,000 person amphitheater, which is an open air amphitheater, huge stage, seating, as well as rooms in the amphitheater tower that overlook the stage. So in many of those rooms, your balcony and your room is your ticket to the show. So kind of a mini luxury suite.
Speaker 1:We also have a the first Hotel adventure park as well as our own lagoons and a water park. So big, exciting pre opening. Thank goodness. Because we're got quite a bit to do to get ready.
Speaker 3:I know even though the property's not open, you certainly are very busy getting the tech stack lined up. You recently published a couple articles on LinkedIn detailing some of the things you've been up to. It sounds like the premise is that you made a bet that you could build a dozen applications yourself. Where did that bet come from, and were you feeling pretty confident that you could make it happen, or how are you feeling at that point?
Speaker 1:Honestly, the bet started with me, but I really couldn't have achieved any of this without a lot of support from by resort. It didn't really start out as a bet. I was playing around with some AI as I was watching vendor proposals. We were looking, you know, especially for internal tooling. I started to feel like the gap between what AI made possible and what the industry was charging for a lot of these internal tools was getting pretty wide.
Speaker 1:So, you know, Vi essentially gave me room to start playing around with items that, you know, were on our roadmap but didn't really exist anywhere. I think that part is important because, you know, AI doesn't make this work, you know, possible by itself. The real key is is a leadership team and an operations team that are committed as well to find a better solution. So I think probably, you know, what we got our executives start asking was, you know, what becomes possible when that marginal cost of building internally drops, you know, incredibly with the help of AI?
Speaker 3:I think in our industry at least, companies always lean toward the buy side of that equation, like when considering whether build something in house or buy something or work with a vendor. How has AI changed that equation?
Speaker 1:Well, I think again for me, I have a background in commercial software. I worked with for a software vendor for years. I started out as a database administrator. I was never actually a coder per se, but I was again a product manager for a while, an implementation director. So I have a lot of experience in all of the different areas of application development.
Speaker 1:And what AI really enabled for me was the engineering team
Speaker 3:you
Speaker 1:know, and the QA team and all the testers. All of that is available to you through these, you know, through these coding tools. Still, think, you know, the hardest part of this is still to figure out exactly what to build and how to get the data, clean data into the system as well. User experience too. So while the AI does, I mean, it's just incredible the volume of work that can be accomplished with these tools.
Speaker 1:Then the issues you run into really aren't engineering problems. More operational problems and understanding just how it needs to function to be useful.
Speaker 3:Well, I would love to hear some examples of apps that you've built and what those specific use cases are. But overall, do you feel like there are some categories of hotel software that hotels probably shouldn't try to build on their own versus other types that are a little more feasible?
Speaker 1:Yeah. Certainly, there are apps that are probably off limits for for internal development. I think that your property management system, your payment processing, you know, your core CRM perhaps, your POS. All of those core enterprise applications are hardened and built with native integrations to each other and I would consider touching those. I think that, like I said, there's a lot of ancillary stuff that you can do internally rather than buying applications.
Speaker 1:My background, I said, was database administrator and BI tooling. So for me, one of the examples was the ability to pull down all of our information from Salesforce. And I got interested in how much could I get out of Salesforce. We've built customer journeys. We have profiles that we're building from our website every day, you know, all kinds of stuff.
Speaker 1:And some of it, I'm not even a 100% sure because we've worked through that over the last few years. So one of the apps I built uses the Salesforce and Marketing Cloud API to connect to both Salesforce and Marketing Cloud, interrogate all of our data there. So all of our client profiles, all of our journey definitions, our unsubscribe page definition and logic, pretty much all of the assets that were out there. And then extended that tool to go and pull it all down, download it all, encrypt it in flight, and place it in our data storage area where I can then go and we can then pull that into our data lake, data warehouse, BI tooling. Doesn't rely on any specific well, relies on Salesforce tools to get it down.
Speaker 1:But once I have the data, you know, I can do it. We've done something similar with Mailchimp as well because we've been merging some different we've used some different tools along the way. So again, said why not And looked up the Mailchimp API and did the same there. Now I can run that regularly. I can schedule it so I can just run it ad hoc and I have complete access to all of that data.
Speaker 3:Do you feel like accounting for your time and any associated costs, like I'm assuming you're paying for some kind of AI tools, is it actually cheaper?
Speaker 1:For sure. I mean, think these tools and this approach has really accelerated a lot of the work I do, a lot of the analysis that I do. I mean, I do occasionally use AI for something besides application development. And I am amazed by what kind of multiplier that is for efficiency. On the other hand, I also want to say that we are extremely conservative.
Speaker 1:I don't know if conservative is the right word, but security conscious. Our VP of IT, Carl Schafer, is definitely security first. And so I want to also just make it clear to everyone. I mean, in order for me to utilize these tools, we ended up creating an isolated development server on a virtual machine completely insulated from the rest of our environment. Could only talk to the coding, the coding tools, and outside.
Speaker 1:I was the only person that could access that machine when we got ready for user acceptance testing. We put it on another isolated machine and only let the user the acceptance test users get to it. Within the applications also, they're they're hardened. And you can ask AI to help you harden it. And then for one of them, we just decided to implement single sign on with our Microsoft Azure environment.
Speaker 1:Carl had mentioned that one of our other vendors, it was going be about six months to implement that with one of our commercial applications that we bought. We turned it around between myself and one of our, you know, network IT guys. We implemented SAML in that tool in about four hours. And so now when you go to it, you just click sign in with Microsoft and it just uses your domain login to get right into the tool. So yeah.
Speaker 1:It's amazing. As far as time, you know, my time commitment to this, I have a full time job. And my full time job is evaluating and configuring and selecting and then working with the team to build our tech stack. I would say that probably I spent about 20% of my time today in application development and in working with the team specifically on these tools. There's a lot of crossover of course because as with a lot of these we're considering tools and I say, wait a second, I can do that.
Speaker 1:Then we and I'll take a minimal amount of time to kind of build a proof of concept and decide, you know, then we decide if it's something we want to pursue or not. So cost wise too, I don't know what the server I mean, we have a server infrastructure. So I don't you know, it's a small development machine for us. I'm on a $200 a month cloud subscription, I think. I got worried because when I upped it to the max plan, I stopped getting nosy, ran out of tokens, errors.
Speaker 1:And I started to worry that first month that maybe I had like turned on automatic add more automatically and I was going to have a $2,000 coding bill. But I feel like I am using it excessively, and I'm good with literally the $200 a month investment in that. Stack wise, I went from using Claude code directly to using Visual Studio with the Clod coding plug in. So it takes less space. It seemed all the cool kids, all the cool coders use Visual Studio.
Speaker 1:So I just felt like that was the right way to go. But again, from what I've seen, it's all the same. You know, the other thing that we've done internally here, I guess it's an expense but it's also a security concern is building our own internal AI models and our own internal LLMs. So rather than because we are even with an enterprise license and all of the guarantees that gives, we we don't put confidential information in AI, at least in public AI. To kind of get around that and to allow our people here to utilize AI with more confidential information, not having to redact it all, we're working at setting up internal LLMs so that we are actually running the AI internally on our own model.
Speaker 1:We train it with what we want. We can configure it. So instead of going out to Anthropix or going out to Microsoft, we're just going to our own internal version of Cloud. It's a little slower because we don't throw the resources that the third parties do at it. But again, it's sufficient for what we're doing and it's much safer.
Speaker 3:Do you think for the typical hotel GM or commercial leadership team, will they need to level up their technical literacy to be cognizant of security risks or, you know, how to think about opening up access to different people. Do you think there will be some upskilling necessary industry wide?
Speaker 1:Absolutely. I think it's just like the PCI rules that everyone had to conform to and all of
Speaker 3:the
Speaker 1:privacy laws. I think that this hasn't caught up with AI. We haven't seen a big breach of any of the public AI yet, not that I'm aware of. If there
Speaker 3:has, they're keeping it, you know. I mean, there are.
Speaker 1:And you've seen like Anthropic had to pull back their latest model because like within three days people, they got it to do things it wasn't supposed to do like jailbreaking it. So yes. I don't know if that means every GM needs to have someone on staff. If there are consultancies that are and I'm sure there are. I think it's a lot of research.
Speaker 1:I think that the standards will be created soon enough. I think that also maybe will help them. I feel like I'm on the trailing edge of this all the time. And I don't know that anyone that really feels like they're ahead of the curve with AI because it's just it's changing so fast. So you keep up the best you can.
Speaker 1:Like I said, I think, yes, you all everyone needs to be AI aware. I think that there will have to be some better standards, better secure access to those security resources. I think it is a big risk. And I guess that's that's that's the counter to all this. All of this new technology comes with new risk, and I don't even know that we we understand it.
Speaker 1:Once your data gets out into AI, how do you claw it back? Or can you?
Speaker 3:Do you have any recommendations on resources or best way to learn about these tools? Like, would you encourage peers to just get a cloud subscription and start playing around and learning it that way? Or is there, like, a newsletter you subscribe to or some way that you've been able to learn?
Speaker 1:There is a new hospitality AI collective or organization that I know, Ira Volk, is very involved in. She's been a huge voice in bringing AI to hospitality. I don't recall what the name of the organization is but it's like the AI Hospitality Collective or but anyway, I think that's probably a great resource. They're all over LinkedIn. Irovolk, I think is probably the person promoting it the most and they've got buy in from a lot of the hospitality vendors.
Speaker 1:And that is probably will be the single best resource.
Speaker 3:Do you think there's a particular category of software that's most vulnerable to disruption by AI or like a part of the maybe buying journey that will change? Or how do you see the software space changing within the hotel world over the next couple of years?
Speaker 1:I think that AI and this ability, you know, again and specifically in this application space, you know, AI will create much smarter buyers. I think that the tools, the research that's available, the analysis of APIs and integrations and existing, all of those data points that can now be combined and fully analyzed and looked at from every single angle are there for you. So in some ways, you know, just being a I think number one is probably just making people better buyers. Number two, yeah, I think those single use, you know, the 75 little applications that you've got running all over the property that you can't even define sometimes, whether they be little connectors that move POS data into the PMS or, like I said, the single I think the smaller single use applications, that long tail is really what's going to be affected. However, I think that, again, a lot of these applications maybe can keep themselves relevant by being more integratable.
Speaker 1:The open APIs are a game changer. Like I said, I just had AI go look at an API. It told me to go set up a key in Salesforce and in in Mailchimp. I went and set up an account, you know, carefully. You didn't, you know thing I learned pretty quickly is you don't you don't paste the key into AI to have it go put it in the right place in the code.
Speaker 1:You actually have to go do that yourself. Because once you do that, your key is exposed. The good thing is that Cloud tells you that and says, okay. Time to go wrote that was great. We're gonna put it in there, but now go rotate the key and do it yourself.
Speaker 1:Those interfaces, those small integrations, but even the UIs are pretty cool. You can customize the user interface and overlay it onto the application that does something that you really need. If you don't like the way it looks, you don't have to wait for the vendor to do it. And so if they have an API, you can create a whole front end. You could do that for a dozen apps and give yourself your own common look and feel even if it's working across a whole bunch of different applications.
Speaker 1:And the more I talk about it, the more I tell people the most dangerous time part of my day is my drive into work. I've got about a forty five minute drive. And in that time, I can like come up with, you know, a dozen new application ideas or enhancements to what we're doing. I'm one of those guys that does like note to self, you know, on the way in. I'm I'm getting pretty good at just doing a bunch of recordings and then because I lose track of half of them by
Speaker 3:the time I get here. Looking back on the past year, what has surprised you the most? Anything unexpected or anything that was easier than what it would be or difficult?
Speaker 1:Yeah. I think there's probably three things. I mean, one, just this assumption that, you know, that operators can't build software. That has been true for, you know, twenty plus years, you know, with the exception of maybe the Excel Workbooks with the macros that the revenue manager has perfected over the years. But yeah.
Speaker 1:So I think definitely in our industry, that's not true. There's a lot of opportunity there. I think too, the assumption that the bottleneck for AI in hospitality is the tech. I think the bottleneck is really around the data and the judgment. AI and broken data is dangerous because it gets you confidently wrong.
Speaker 1:You know? I'll give you an example of that. I did one of the other apps we're building internally is is to manage our SOPs. And we're building a lot of SOPs and we started building them all in Word documents on the templates. And I thought, wouldn't it be great if I could pull those all into a database and make them accessible via conversational AI?
Speaker 1:And so just went off and sucked all the data in, know, had AI do whatever wanted to to get the data in there and I started trying to talk to it. And it would give me kind of the right answer usually with chunks of other bits of information all over the place. So what I realized was it's because we're not consistent. You know, the the way the SOPs were written, although it followed a template, they weren't necessarily even procedures. You know, they were statements.
Speaker 1:They were other. So the goal actually turned out to be to create a tool that would standardize the written SOPs. And so we fed it with kind of the voice of I, you know, how we wanna talk to our employees, what our standards and our pillars and and all of that. We said, you know, for different departments, there are certain guidelines, you know, on how F and B does things, how, you know, procedures around HR. And so we built all of that structure in and then we're pulling all of that data through this tool into a database now.
Speaker 1:And it it structures all of the SOPs into actual procedural lines where it read an SOP and it said, you know, this is actually three separate SOPs. It splits it out. It presents it to the person who wrote it with what they had, what it rewrote, and then even gives them a little summary of why it made the changes it made. They can edit it, accept the changes, say, no. They keep it the way it was, save it, versions it, puts it in the database, and makes it much more accessible and usable by the AI.
Speaker 1:So the data, I mean, it really that was all that's just been a really long way to say that your data has to be clean. It has to be well structured, And that's just gonna become a bigger and bigger requirement through all of this. Again, probably the AI will help you do it.
Speaker 3:I wanna wrap up with one last question based on everything you've learned. If you could give a hotel GM or CEO one action item to do over maybe the next few months, what's one thing you would recommend to help them learn or implement AI in their businesses?
Speaker 1:I think learning AI, there's a lot of opportunity out there. Again, I have an application development background. So so I kind of dove into it from that perspective. And I actually am, you know, 99% self taught. There are so many organizations, so many resources.
Speaker 1:There there's a lot of courses on AI. I applaud specifically, but on all of those, I would try and educate yourself on whatever piece of that is of interest. You know? I think getting the right people involved, you need somebody who who probably understands data. And like I said, I think the data foundation, which is the unglamorous part of all of this, is is really what you need to get your arms around.
Speaker 1:Clean, integrated, properly governed data is key. As far as the resources, I think a lot of people have resources on staff that can start. So it could be, you know, an operator GM that likes to code a little that started playing with Claude, be a revenue manager with those amazing spreadsheets that, you know, wants to do it differently. You know, you get somebody you know, Python is a great thing to learn. You know, you can probably get a basic education of Python, which is probably one of the more popular languages.
Speaker 1:And, you know, I think you probably also have to be willing to pay people to do some of this. You know, I think you could, whether you do it yourself or you wrap it into someone's description, you know, I think that giving somebody who, again, maybe is is technically adept, you know, the opportunity to learn some of this, go take a quad code class, a general architecture class, something, you know, that helps. But, again, whether you're trying to get to the point where you're actually creating your own apps, you want to figure out a different way to to look at and present your data, or you're just playing around so that you have a better understanding of what you're doing. And those are all great, great goals, and I think it all, you know, gets you to a better place.
Speaker 3:Well, thank you so much for your time, Bill. Thank you for sharing your wisdom. Hope you have
Speaker 1:a
Speaker 3:great rest of your day.
Speaker 1:Thanks. You too.
Speaker 2:That's all for today's episode. Thanks for listening to Hotel Tech Insider produced by hoteltechreport.com. Goal of this podcast is to show you how the best in the business are leveraging technology to grow their properties and outperform the concept by using innovative digital tools and strategies. I encourage all of our listeners to go try at least one of these strategies or tools that you learned from today's episode. Successful digital transformation is all about consistent small experiments over a long period of time, so don't wait until tomorrow to try something new.
Speaker 2:Do you know a hotelier who would be great to feature on this show, or do you think that your story would bring a lot of value to our audience? Reach out to me directly on LinkedIn by searching for Jordan Hollander. For more episodes like this, follow Hotel Tech Insider on all major streaming platforms like Spotify and Apple Music.
