Your Hotel’s AI Problem Is Actually a Data Problem

Independent hoteliers are under pressure from every direction, and now there’s an AI mandate on top of it all.

Wil sits down with Adam Harris, Co-Founder and CEO of Cloudbeds, to cut through the noise. Adam’s argument: most operators aren’t failing because they lack effort or the right tools.

They’re failing because they haven’t defined the problem, and they’re sitting on fragmented data that makes even the best AI useless. The takeaway: AI isn’t the strategy. Better questions, better data, and better decisions are.

This episode is presented by ⁠⁠Cloudbeds. Connect with Cloudbeds at https://www.cloudbeds.com/gmh, and you can subscribe to Adam’s newsletter here: https://www.cloudbeds.com/newsletter/

Watch This Episode

Today on Good Morning Hospitality, I’m sitting down with Adam Harris, the co-founder and CEO of Cloudbeds, one of the most widely used hospitality management platforms in the world, serving independent hoteliers across 157 markets.

We’re getting into the real pressure independent hoteliers are feeling right now, with rising costs, fragmented tech, distribution complexity, and why the AI conversation most operators are having is actually the wrong conversation.

So if you’re trying to figure it all out or where to even start, this is the one for you. Let’s get into it. Adam Harris, long time, no see, my friend.

It has been, what, five years-ish?

It can’t be five years, has it?

No, no, we’ll just say it hasn’t, and pretend like, you know, it was five days, five weeks-ish, five months.

Too long, too long. Too long.

Too long, well, for the listeners, it’s good to see you as well.

For the listeners of Good Morning Hospitality, they may not know, I actually had you on Slick Talk, my original podcast back in the COVID days, where obviously hoteliers, STR operators, everyone was just trying to figure out what to do.

1:20

Apple Podcasts

And now here we are kind of in a similar situation, just maybe not as chaotic and hectic in terms of AI and data.

1:37

The Operator Squeeze

So Adam, I’m really excited to jump in, and I want to kind of just push in right out the gate on the squeeze the operators are really feeling. Hoteliers, I know you hear it, we hear it, the rising costs, more tech, more distribution complexity.

And now, the last two years, basically, I’ll give two years as our cushion. The last two years, everyone is telling them to use AI and that they need an AI strategy while they’re just trying to float and survive.

So the question I want to look at with you is, where are owners and operators truly feeling the pressure?

Because you at Cloudbeds all hear every hotelier’s problems, everything that’s going right, maybe not as much as the problems, but yeah, I would love to get your take, Adam.

Well, look, I’m in a really unique position. I get to talk to thousands of hoteliers a year, which is an absolute blessing. And the common areas without getting into specifics is, I’ve got a lot of cost pressure.

I’m dealing with fragmented tech. I’m challenged in how I think about my distribution and my strategy and how that’s evolving. The guest’s expectations are moving faster than my staff and I can keep up with.

And the reality is this, hospitality is incredibly hard. It is one of the grittiest industries I can imagine. And frankly, I don’t think there’s another industry where the person that you’re serving spends the night with.

I think it’s medicine, it’s hospital, nurses and doctors have that same behavior. And so hotels do that 24-7, 365 days a year. Look, so yeah, you guys wrote an amazing article on the squeeze.

And I think what really sort of reinforces there is like, the expectations are the same, but all the inputs for those expectations around the guess, as well as trying to maintain profitability, are being exceptionally challenged.

And here comes this new technology that’s supposed to make things easier. And it does and it doesn’t in some circumstances. And so I feel for the hotelier these days.

I really do and that’s why we’re trying to build a system that brings as much of this into a consolidated area to streamline costs from a tech perspective, but more so help free up some of their staff members to focus on the guest side of the arena.

So lots to talk about there.

Yeah, definitely.

4:04

When Tech Becomes a Burden

Adam, as I’ve been listening to you guys talk about, going away from a property management system to a profitability management system for hoteliers, and as a tech CEO, I want to know at what point does the tech actually start to help?

And when does it become part of the problem? When is it too much tech? Because I think that’s where a lot of operators are feeling today too.

The operations are getting more complex. A lot hasn’t changed, but maybe the pressure has changed, the expectation, all the things that you’re saying around the guest. So I would love to know your perspective.

So this is a great question and I’ll sort of follow it up with some questions.

Is this a tech problem? Is this a distribution problem? Is this a profitability problem, meaning how we operate hotels in this day and age?

I think it’s the independent hoteliers that I serve, and we’ve got 20 something thousand of them around the world. And so it’s not just in the US, it’s literally around the world in 157 markets. None of them are lacking effort.

Let’s be very clear, right? But what they are lacking is leverage.

And so there’s an extent where too much tech becomes burdensome and too little tech removes some of the gains that you might have with the right balance of technology and human involvement.

And so what I know from my thousand conversations with hoteliers is none of them wanted to get into this industry to deal with technology. None of them wanted to be a professional marketer, right?

But what they did want to do is show off their art form.

And so when I think about that art form, which is taking care of gas, which could be a certain concept, it could be a wellness concept, it could be a limited service concept, in everything in between, managing 19 systems, which is the average number

of systems that the average hotel deploys is not the right answer. But when you think about where leverage comes into benefit with tech, it has to be solutions that simplify their day-to-day, right? Give them time back.

It needs to organize data in new ways, and then it ultimately needs to furnish how they act faster, and they make decisions and they solve for solutions. That should not be an added cost. That should be part of the service bundle.

And I really don’t see that happening today in this travel ecosystem. I see a lot of promises with very little foundation.

I see a lot of shifting in brand positioning, going away from, hey, bring what you want and tie to my PMS, to we’re now an operating system or a hotel management system or whatever. And that’s normal, right? Like that’s going to happen.

We’re seeing a regime change in technology. But the best part about that regime change is no one owns the monopoly to it. You and I can go and spend $20 a month and get an access to the best AI models on the planet.

And we can use it for our own lives every single day.

And that gives me a lot of comfort because as we all get more familiar and as we sort of catch up to the latest phase of this AI evolution, we can start to implement solutions and get some real benefit.

But that’s not a dive in head first into the deep end.

7:25

AI Starts With Questions

It needs to be methodical. And so what most of my conversations have shifted from of how will AI help me, Adam, to what are you trying to solve for? Let’s work backwards.

Let’s work backwards from the solution to what are you ultimately after?

Yeah. Well, it’s funny because even during our pre-chat about this episode, you’re talking about operators don’t even know what they’re trying to solve for because they’re just trying to survive.

And I think when you’re in that survival mode, it’s really hard to step back and think about that. So where is AI actually useful today that will still apply as newer models come out, as it gets better? And where is just noise for you guys?

Because you’ve even been playing with agentic AI for a couple of years. You and I have chatted about the deep dive that you’ve gone into personally on AI itself. So as much as I love what we’re talking about, it’s hard when they’re in survival mode.

And so I guess where do the independents have the leg up with this now?

Well, there’s no brand standards that they’re going to have to follow. There’s no systems that they’re required to enact. Some of those systems can’t even do the agentic harness layers that are determined to be deterministic.

8:37

Deterministic vs Probabilistic

So I want to talk a little bit about deterministic and probabilistic. The data is a real issue in this industry.

If you have data across 19 different systems, you are having to create some form of centralization of that data to optimally enact a decision from AI. And then you come into what is like the natural building blocks of AI.

It’s to sound very human, right? That’s why we use them because we find them relatable to some degree. You ask it a question, it gives you a pretty good contextual answer.

But that is built on what’s called a probabilistic engine.

And in the best way I can describe that is how many of us listening to this podcast, or you and I included, started writing a paper at one point in time on a computer, and the power went out and you forgot to save it.

You’re like, I got to start all over. I mean, I remember those college days. It’s the worst, but you do it a second time.

The context is all there, but it’s slightly different. Well, that’s the same thing that LMS do. If you took 100 inputs of the same question, you’re going to get 96 different ones.

Four of them will be the same or roughly the same. Deterministic means you ask the same question at 9901, 902, 903 throughout the course of the day. It always gives you the same datapod back.

Well, in hotels, occupancy is occupancy. A rate is a rate. A I have a room or I don’t have a room is deterministic.

It’s not probabilistic, it’s deterministic. So we need to solve for the data side first, and then it’s really easy to go, how can AI help me with marketing?

How can it be part of my guest communication or my hyper-personalization engine or revenue or service recovery or whatever? It’s great at all those things.

When and if it has a deterministic datapoint and it has a deterministic model helping them seek those answers, I spent two years with Cloudbeds solving them. This is not something you just turn on with the latest OpenAI, the latest Cloud version.

It is all in the data orchestration. It takes tremendous amount of compute processing, and we’re very excited about that next chapter.

And other people will get there too, but the industry in a whole has to solve the data issue first, and then we can evolve to where this is the most applicable for my need.

But I don’t know a single marketer at any hotel that can write 12,000 personalized messages to every guest that has ever stayed with them. That being said, all the data is there to do it today.

So someone just has to go physically, manually write 12,000 personalized emails. Well, AI and a swarm of agents can do that in five minutes.

Yeah, no, I love it. I love it. It’s really, it does make me reflect on what we’ve been doing kind of internally because as I’ve been looking at our media side of things, right?

Like all of our data is on YouTube or on Apple, on Spotify, all these different platforms, right? And how do we know that we’re actually serving our audience better with a proper podcast or how we’re framing things?

And that was kind of my hassle because I was getting, I had a great strategy.

I had a mess of data and that was kind of like where I was getting a poor output from AI versus now when we’ve cleaned up our data, we’ve organized it, we’ve centralized it.

And that looks different for us, obviously, because we’re a media company, not a hotel, but it was night and day difference on the output.

And so you can have poorly organized data with the best strategy and get a pretty not good output from AI, but you can have the most clean data or even just 80% there with a mediocre strategy and have a really good output.

12:24

Cleaning Up Hotel Data

And so I guess how do hoteliers orchestrate their data in order to actually start solving problems?

I would love to know just from even a development standpoint or a product standpoint at Cloudbeds because I remember being my old hotelier days where property management softwares were not the best.

So yeah, catch me up to maybe like where the tech is at today.

Yeah, I mean, look, from Cloudbeds perspective, I mean, we’ve got this salt, right? We have some of the best APIs in the industry and we break it out in three separate sections deliberately. One would be how you sell your rooms online.

That’s all your distribution, the whole guest layer. So the whole guest experience and guest interaction layer and then ultimately your property data layer. We are a centralized ecosystem.

We’ve been unified from day one. We were the original operating system or hotel management system or what do you want to call it? It doesn’t matter.

But the point is, we never saw ourselves as a PMS. PMS is 25 years ago, put our industry in the hole that we are by only doing a narrow thing when they easily could have done a broader thing.

So just from a data perspective, allow everything from occupancy to tax, to inventory all comes back to a human, a human stay. So data is very human in this industry. I’ve said that in the past.

I’m going to continue to say it, but how we organize around that entry point is everything. So Cloudbeds has that solved, AskSignals is one of our newest products.

13:49

Cloudbeds Signals In Action

We showcased this last week with our product release webinar and for our spring releases. And it is an orchestration layer that sits on all the Cloudbeds data.

You can ask it questions, you can get help from within the system of like, how do I do something in the system or hey, how should I plan for housekeeping next week? And it is all deterministic.

It shows where the data came from, how it was recommended. And then adding a little bit of contextual layer is the fact that we’ve trained these models on the hotel industry. It is a hotel expert.

There are different personas inside of this engine depending on who’s asking. Is it a housekeeper asking? Is it a GM asking?

Is it a revenue manager? So like just changing the contextual relevance slightly helps get a better deterministic output. So I’ll give a great example.

Hotel 1550, a property in the Bay Area uses Cloudbeds. They left one of our big competitors who likes to hype themselves. And prior to that, they were a brand.

So deflagged, went to a competitor, left the competitor, went to us. And all of a sudden, all their data was in one place. Their distribution strategy shifted.

They started buying ads. They started building landing pages. Everything that they wanted to do but couldn’t because of past systems was eliminated.

And so now they have Taiwanese custom landing pages aimed at an international traveler coming to the Bay Area.

They’re doing parking packages for the business traveler who doesn’t want to, needs to rent a car but doesn’t understand the complexities of the Bay Area and parking situation.

And so all of these very hyper specialized strategies were furnished and harnessed from our signals capability. So that’s just like a small introduction of something that we’re seeing because the left hand is talking to the right hand now.

It’s seeing operations and distribution as core to how do I maximize my total rev par. And that is something that systems in the past just didn’t do. They weren’t able to get to the root numbers and understand the building blocks from that.

And now we’re using those building blocks as recommendation tools. And so I’ve never been more proud of that team that built that product. The fun thing is the agentic side of it was relatively easy to do.

Like that came really quickly. But the data harness, wow, my goodness, that is complicated.

Yeah, I was going to ask what made it so easy, but I guess when you add the data component, that answers it for me.

16:35

Autonomous Coding And Caution

That and look, I mean, one of the fastest growing areas is the ability to leverage these amazing tools to code.

I mean, we run 24-7 autonomous agents writing code. We have been for years. So these tools, we can give them a project and say, here’s the delivery.

And they’re like, oh, but here are all the endpoints and all the data sources that you need to connect to. And it’s like, yeah, it just writes the code.

And it gives me a lot of hope for not only our industry, but also Cloudbeds customers, because we are deploying more and faster than we ever have as a business, because we want to bring these tools forward to our customers.

And fun fact, coding is basically free now, but yet there are more job postings for engineers than ever in history, right? So that goes to show you how important it is. And here’s the other comment that I’d make.

This problem, this data problem, is every business’ problem, from Metas down to Google to Apple to the smallest companies out there. And so like, we are all going to solve this. We are all running at this.

We are all learning in motion. And there’s no right answer yet. That’s why, don’t dive head first into the deep end.

Like work with brands like Cloudbeds, who have a very good focus, that are talking about deterministic. You hear that in my post. I’m not reading that from any other brands, because they have not solved it yet.

So until they start talking about it in their lens, it means that they haven’t solved it. They’re still trying to overcome that challenge. But look, this is a big bright world that we live in, and I’ve never been more excited from the tech lens.

The squeeze is real, and I’m hearing that, and I hope we can shift some of that, sir.

18:21

Future Of Hospitality Systems

I love it.

I love all that you said. Before I close out with one of our closing questions around what operator should really focus on now, which is obviously a hard question with how fast things are going.

But I want to just ask you, how does Cloudbeds role change as operators become more than just a hotel and more than just a bed? Do you guys see this continually to evolve what Cloudbeds truly is as a tech software company?

Or do you think this era that we’re in is going to last for a little bit longer?

Oh man, that’s like the trillion dollar question. Well, look, we’re sitting on tech that we haven’t released or haven’t even talked about because our operators are not ready for it. So we need to start with better questions.

The hoteliers need to understand what their problem is and define it. Then we’re going to organize the data on their behalf. And then we can start to throw solutions at the problem.

I think lodging will change forever. Partially because the expectations as guests will change.

More so, I think if the world is going to be more digitalized and we will be relying on our alter ego in the digital format to do things to streamline it, experiences are really going to matter. It’s really, really going to matter in our life.

And I’m a big fan of Ready Player One. I love the book. I love the movie.

Book more in the movie. But like that world where you’re like going to live your experience, I think is very far away.

However, when I was inside OpenAI’s office, I watched in a matter of minutes a new model, like build Minecraft that was playable in like minutes. And I’m like, ooh, okay, that’s a little crazy.

And if we can build these worlds and experiences in a digital framework, maybe we do accelerate some of where the world goes to. But I don’t lose hope in what hospitality represents. I’m a big believer that we can streamline operations.

We can leverage tech to harness the right format, the right style, the right voice to reinforce the art form. We can give back more time to the operator. And all that’s coming now and coming soon.

So I think from a systems perspective, they’re forever changed. That train has left the station. The operators, once they define the problem and the solution providers meet the answer to that problem, we can then go tackle the bigger issues.

And that’s just how rooms are sold. That is a big expense to this industry. It’s been disintermediated.

It can be further disintermediated in a positive way. And look, at the end of the day, no brand, no OTA owns the rights to this technology. It is a completely free and untapped and unharnessed technology that any operator can live by.

But that only works if we get the foundation there first. So let’s go ask great questions. And if any listener wants to ask me questions, I’m around too.

I’d love to be a part of that discussion more. My newsletter is trying to talk more about this topic as well. And so thanks just for having me.

This has been fun. I think we could talk about this for two hours.

Yeah, I was going to say we’re going to have a lot more conversations in the future because, like I said, this is all moving so fast that any listener right now as an operator, I bet in a year we’re going to be having somewhat of a different

conversation because a new problem has arised or some kind of new thing has unlocked due to centralized data, all the good stuff that we’re talking about and asking better questions. conversation because a new problem has arised or some kind of new

21:55

Closing Takeaway

thing has unlocked due to centralized data, all the good stuff that we’re talking about and asking better questions. I just love that statement that you’ve given us in terms of just like starting with better questions, defining the problem better,

not just saying I’m using AI because everybody’s using AI. As my mom would say, if your friend jumped off a bridge, would you? My answer to her is, I probably jumped off the bridge first. But yeah, it’s really great conversation.

Thank you so much, Adam, for having me. For the listeners, make sure to grab the links in our show notes. Adam’s newsletter is actually going to be linked in there too because I’m an avid reader.

Until next time, we’ll see you guys all again on the next episode.