Mews on Muse: What Consumer Adoption of AI Agents Means for Hotels - Richard Valtr

Josiah: Richard, it's great to see you. Thanks for making time to record. I always enjoy our conversations. I feel like I learn a ton. We're going to talk about a lot of things in this conversation. I want to start out on a note that's maybe a little bit lighter, and it feels a little silly, but with the launch of Muse by Meta, it's been interesting to watch some of the consumer response to personal agents and things like that. I feel like it builds on themes that you've been talking about for a long time, so it's funny to see this play out. But take us back to the origin of Mews. We've talked about this from a few different angles, but what's behind the name Mews, the real Mews?

Richard Valtr: The real Mews is really just the way that I was thinking about it. And it's so funny, because it's such a topical thing. When I was thinking about what it is that I'm trying to build, it was an operating system, and the operating system that still is basically the main operating system of today's world is Windows. But when I was thinking about that, it was like, we're not actually operating in a way where there are all of these different windows that you go through. It's going to be a series of these little departments. So actually they're much more self-contained. You've got your sales department, you've got your marketing department, you've got your revenue management. And so I just thought about it as a collection of little houses.

Then when I thought about it, it's like, oh, that's kind of like Monopoly, where the little houses come together and they become a hotel. And having grown up in England, what's a row of little houses? It's a mews. You've got your mews houses. If you live in London or if you live in New York, or basically anywhere where the Brits spent a little bit of time, you end up seeing these little rows of houses. They're converted stables, basically.

So I always really, really liked that name, and somebody at some point told me that it should be a small enough name for these things. So that's really where Mews came from as a name. Back in those times, I thought it would also be quite cute to have little houses represent the logo and stuff like that. We moved away from that, but it still sticks and it's still exactly that. I always really liked the interplay of the word mews with the actual meaning of the word muse. So now it's quite funny that you've got this Mews and Muse connection, which I just think is a 180 turn on what exactly the world is coming to.

I don't know exactly, but from the vantage point where I am today, I think that personal Muse to your corporate Mews connection is the thing that the world is slowly but surely careering into, where the need for the middleman in that transaction just becomes moot.

Josiah: It's so interesting, Richard. It's one of the reasons I love talking with you. Things that we've talked about years ago have played out. The last time we recorded was December of 2025. We're recording now in October 2026, and it feels like just recently these personal agents, whether it's Muse by Meta, whether it's Clawdbot, whether it's Dot, I feel like every frontier lab has a version of this. And it does feel like that direct connection, that ability to take action, to have agency, to get stuff done, is becoming a consumer expectation. And I feel like those easy ways of getting things done are what you've been building towards for years to empower people working in hotels. If I recall correctly, it was built on this idea that the line between consumer tech and business tech might get more blurry. It's not like we're using tech from the 1980s.

Richard Valtr: But I think that is quite interesting as well, because of the way that we thought about design from a very, very early age. We thought about this from the perspective that hotel systems shouldn't be enterprise tech. Because the people using them don't want training. On a high level, I think training is good when you're training the way to behave, the way to look after your guests. But the actual use of the system, from a corporate design perspective, how much of the time do you want people to be filling in the functions of the system versus spending time with your guests?

I think the answer should be that most of the time should be spent with your guests, because that is the service. That is the thing that ultimately you as a business are getting paid for. So I always think about it as: the systems that you use in hospitality should look like TikTok.

And that's why, for example, even chat interfaces... I think chat is a really, really good interface for a strategist or for an engineer, or for somebody who's putting the concept together. The way that we use, for example, OpenAI or Claude, the biggest use case is coding, or strategizing about your court case and stuff like that.

But those are builder functions. Whereas the dominant display for the way that you do your life on the internet is actually very, very clean, very scrollable, one or two clicks, very easy to action things that are more interactive in that way.

I always think that in hotel tech, we also have to be designing for our customers as consumers, rather than trying to design for them as programmers.

Josiah: I want to pull on that thread for a moment, because I enjoyed seeing you on stage at Destination AI recently, and I feel like this was coming up a little bit. From our conversations, from watching you build, you're building the future that is possible, and you're also attentive to what is unfolding. As it relates to either the associate experience or the guest experience, how do you think about directing it versus only responding to what people are asking for?

I could see both being valid. I know you've talked in the past about this future that we could have. How directive do you want to be in creating that future and creating the technology that enables it?

Richard Valtr: I do think that the biggest change that will come to hospitality is not necessarily us having an opinion, but actually the data having an opinion.

That's one thing that, even for most hoteliers, I think is going to feel pretty jarring, because most hoteliers don't want to be told what to do. For example, we will present: this is what you can see as your prices, and these are the prices that you should get into. And I think about those prices, the governance model for how you set the prices. I always think of my arguments with my mother. At one of the hotels that my family had, at some point I was the revenue manager, and she would say, "Well, look, we're a 250 euros a night hotel. I don't want to see you setting the rates lower than that."

Or sometimes she'd be like, "Oh my God, you're selling this room for 650 euros a night. How?" And I'm like, "The maths has guardrails, but it doesn't have an opinion." So on this Tuesday night, you might feel that you're a 250 euros a night hotel, but in this market you're a 120 euro per night hotel.

And maybe, okay, if you feel that, let's drill down into it. Let's test that hypothesis. But I might need to reach the revenue goal that I have for that day. I might need to sell this hotel at 120 euros, because that's what the market is telling me.

I would love to sell it for 250, but we're not going to reach any of our daily revenue goals selling it for that. Now I can think about: am I speaking to the right audience? Am I making sure that the types of customers that I'm trying to get for that day are there? Have I signed the right corporate contracts?

All of those things are super, super valid. Maybe if I'm going on Booking.com, I shouldn't be selling it there, because that also persists, and then you get those weird things where it says this hotel is between 120 and 850, and that also sends a weird message to the customer.

But again, there are things which are maths-based and there are things which are just feeling-based, and I do think that there is more scope in this industry for us to be more mathematically and data-based than "I know my business really, really well." And I think that most hoteliers are actually responsive to that, and they want their systems to tell them, "Look, this is what the data's telling you. Now you have all the power. You decide."

But I think most of them do want an active role from their system, to tell them, "Look, this is what the data is telling me for this thing, and this is where it's pointing to. Now here are three choices of what you make of that. This is probably the impact that those choices will have. Let me lay it out for you." Whether you like it or not, or whether you think that it destroys a bit of the autonomy of the owner or of the GM, I think that's where we are today with technology. On the one side, that is what the market is telling us that we should be doing, and on the other side, it's where technology is moving to.

And the counterpoint to that as well is that you just can't be a PMS anymore. Because in order to make that decision... I just described to you a conversation that I had with my mother about this, but if I talk to you in terms of systems, then I mentioned something about the corporate rates.

I mentioned something about the segment of those corporate rates and the types of companies that I should be going out to. So that's your CRM question. I also talked about what rate was being set at the time. That's your RMS question. Then you've got the way that interacts with your PMS. It's also a distribution question. Am I presenting those rates on Booking.com, or am I only doing it through private channels? Is it a direct connection that I have to Navan, or to Revolut, or my JPMorgan account, or is it something else? And so I think that this is the other thing. When people go, "Well, companies should stay in their lane and make it open."

But the problem is, intelligence is not open. Intelligence is the whole thing. And so having an opinion just as a PMS on what is the next best action, it's now too difficult to do that. And I think to our credit, we've really been the one company that's really tried.

For example, five years ago, for every single RMS, we made it fully dynamic. So if you wanted to add a breakfast to the room price, you could set that dynamically. One of the reasons why we had to go out and get our own RMS is that we just thought that in the future, especially when you're going to be trying to do dynamic segment pricing, or the furthest route that you can think about, which would be dynamic personalized pricing, or loyalty pricing, or membership pricing, or whatever it might be, then you need to have every single one of those different things. Whether it's the F&B component, every single aspect of that 72-hour journey should now be dynamic and respect those conditions.

That's essentially what the intelligence layer should be asking you, or should be telling you. It's no longer just the room.

Josiah: Yeah, it makes a lot of sense. I want to come back to technology architecture and strategy, because I think this is really important for hotel leaders to think about.

But one of the things we were talking about off air that might be nice for a little more context here is you had this notion of the thinking hotel. I wonder if you could describe what this is, what it means, and what it might present as an opportunity in the future. And then I want to come back to architecture, because I think it will support this thesis you have.

But first of all, what is the thinking hotel? When you think about the thinking hotel, what is this opportunity, or this concept, that you see?

Richard Valtr: I think we spend a lot of time thinking about what it means to have a system that has intelligence, or what it means for the hotel to actually have intelligence.

On the one side, it means that your knowledge base is consistently being updated. So now your knowledge base isn't just the thing that you put on your website. Hopefully you put more information on your website than you do with Booking.com and Expedia, but that's not always the case. But it's the thing that presents, because nowadays you'll have a Muse agent, M-U-S-E, that for whatever reason has picked out your hotel from all of the different hotels, and it now wants to go and find all of the information that further confirms it's made the right decision in recommending that this hotel is one of the five hotels that matches the selection criteria of the guest.

So the guest has said to his personal agent, "Go and find me the best hotel, and I have these selection criteria that I want it to hit." And so now your knowledge base becomes a really, really necessary component of that, because now somebody is trying to communicate with your knowledge base directly. And so that knowledge base is always going to be updated, because it's every single text that is shared between you and a customer.

It's every single phone conversation that you have with a customer. It's every single email conversation where somebody has replied. But it's also every single AI interaction, if you have, for example, voice AI or messaging AI. And the hotel is constantly trying to learn more about what people, whether it's real people or whether it's agents, are asking.

And so you need to constantly be updating that hyperlocal semantic layer of the hotel, so you have a knowledge base that can answer all of these very, very complex questions. And to a certain extent, we've always known this in hotels. What do you do after you hand somebody the key to their room?

You go, "Well, the elevators are there, and checkout time is 11 a.m. Breakfast starts at 7 a.m." That's the thing that we say. And now what I think becomes better is that the associate who delivers that can now be prompted: that person who's checking in actually texted you on WhatsApp, and the AI agent already told them all of those things.

So here's a bit of information that the guest doesn't know because they didn't ask. So maybe part of that is going to say, "And here are three other facts that you can mention to the guest that they don't know yet."

Josiah: Interesting. And that's how you add value in this world where everybody's plugged into everything and you have seemingly anything on demand.

What is the proprietary information? What does surprise and delight look like in this world?

Richard Valtr: And that's what I think about. Now it's an organic process. It's the same way that we as humans interact. And you can slightly bristle... I hate the anthropomorphic aspect of AI.

It's so funny, there's a raging debate within Mews about whether we should call any of these things coworkers and stuff like that, and I've just banned it everywhere. I'm like, "I do not want anyone to start telling people that these are real people working."

Josiah: Yeah, it's software.

Richard Valtr: It's software. It's a tool. And I think it's important, even though people are like, "Well, the customers know this, they're not idiots." But it's not about that. It's about showing what side you are on. Are you on the side of an Anthropic, where you think that you're building a god?

And this is some kind of smarter thing that will now tell you everything. Or are you actually building something that's a tool, and you're saying that it has to be incredibly helpful every single time? I think that's the way that we try and think about it. And so from that perspective, you have to think about how we're going to be helpful, because the hotel itself is now part of this super intelligence structure.

In a way, it was always part of a super intelligence, but now it's part of an even more rapidly learning structure that says every single interaction is going to be more intelligent than the one the day before.

Josiah: I think this frame also makes AI more useful, from my perspective. We were talking before recording about this idea where software could be the harness, I think, as you were putting it, to deliver guest experiences.

But that only fundamentally works, I feel, if there's some guidance and control and strategy, as opposed to just thinking this is going to be probabilistic, "Let's see what the AI comes up with now." It's very hard to build a hospitality strategy based on chance.

Richard Valtr: Yes. And I think the one place that we're in right now is a very strange place for us to be in as humans, because we know where we are today. Still, when I speak to Claude or OpenAI or whatever, I'm still like, "Oh, he's an idiot," or "She's an idiot," whatever. And I'm better than that, or I know more.

It knows more about the things that I don't care about, which is great for learning, or that I hadn't previously cared about, so I can learn more. I'm not specialized in all of these different things. I don't know every single statute of GAAP accounting and all of these things, and now I can just ask something that is basically that, and it can tell me a reasonably good answer.

But about all the things that I do know, I'm still better. And I think we're in this weird place where suddenly people are getting worried that maybe this thing will replace all the things that I bring in terms of value, and maybe this thing will start doing those things for me.

And I think it brings about this great crisis of confidence that people have, and I'm weirdly not moved by it. The way that I see it, and this is something else that we talked about last year, there will roughly be two economies.

One is this machine economy, where machines talk to each other, probably in maths. The fact that they currently speak to each other in human language is actually really, really useful. For example, if you want a chain-of-thought analysis, it's actually very, very easy for you to re-engineer how a particular thing has come together, because the model will tell you, in a probabilistic way, how it got to that.

But generally it'll tell you how it got to that thing. Whereas by the time it speaks in maths, it's going to be really difficult for us to do that. But in my mind, we're a year or two away from that. And there is this other economy, which is the human economy. The machine side will be continuously more and more theoretical, and we'll see it in terms of progress.

We'll see models being proved. We'll see theorems being offered up as proof. We'll probably see mathematics as a profession become less and less attractive for people to go into. But generally, in the human professions, what we'll see is actually a huge explosion of human professions, because we'll be able to dump so many things that I think humans don't really enjoy.

There will be a machine economy that will increasingly get cheaper and cheaper, but more and more sophisticated on that side. And there will be a human economy that I think will also get more and more interesting. And the reason is that we will have more ways to feel good about every single one of our decisions, and the quality of our own thought will improve.

I don't think the act of being a hotelier, or of providing a hosting service, will become less enjoyable. In fact, I think that it will become more enjoyable, because we'll just be able to do more. And there will be an abundance in the human economy based on these things, because I think the machine economy, and this is me being really, really Pollyannaish, will get us to cheaper and cleaner forms of electricity, because the machines will figure that out for us.

I don't feel we're that far away from fusion now because of this. I don't feel that we're that far away from being able to harness the sun. We no longer need to wait for billions of dinosaurs to die for us to burn them and destroy every living thing in the process in order to get the electricity, because we'll be able to take it from the giant fusion reactor that is the sun.

Or, for example, make some of our own, because all of these things were always thirty years away. Now I feel like some of those things, because the cycle times are going to get much more compressed, will be much closer. And I think that ultimately the human industries, of which hospitality is the peak, will really, really gain from this.

Josiah: It's interesting to think about that in a world of abundance. If things are faster and cheaper, what becomes valuable? And so much opportunity exists within hospitality. I would love for our listeners to get a bit of a sense of what it looks like to have technology to thrive in this world.

Where, again, we want it to be human-forward. But you mentioned a little bit ago that intelligence can't truly be open, and at Destination AI you had this analogy that I found powerful. I think you were using the iPhone, or the Apple analogy, when you were talking about open versus closed systems. I think this is something that many hotel leaders either are thinking about or should be thinking about.

My question for you is, what do you see as the role of Mews in this future, as you build it and as you watch this stuff unfold?

Richard Valtr: I look at it as the same as what it's been since day one. There's a reason why the design of Mews looks like the design of Mews. I fundamentally think that it's a little bit old now, but the thesis still holds up. The way that I designed Mews was the general manager's view.

You want to know what's happening at your front desk, you want to know what's happening with your reservations, you want to know what's happening with your distribution, and you want to know what's happening with your finances. So you have that total view, and then it splits down.

Ultimately, a front office person should see their front office work. The reservations department should see their reservations work. The revenue department should think about their revenue things. But I think what will change is that we will not see it as disparate systems.

"Oh, I do my revenue decisions in this system, I do my distribution decisions in that system." It'll just be a series of these cockpits, which give you that view. That was the whole reason why we did the open API in the first place, and why we did the marketplace.

And the way that we always thought about that was that it had to be two-way. We were so bolshie about this at the beginning, because we told everyone it had to be a two-way connection. And this will get really technical for a lot of your listeners, but we said whatever you're seeing in Mews has to be the same thing that you're seeing on the other side.

So in the end, it doesn't really matter if you're using the Mews interface or you're using a different system. You should know, and be happy about the fact, that those two systems are perfectly interacting and talking to one another. To me, we are still keeping to those same principles.

And I think the only thing that changes from this perspective is that you now have systems where that talking time needs to be massively, massively sped up. But from that perspective, I don't ever want to say to anyone that they only have to use Mews.

The analogy that I shared was that it should work the same way as when you're using your iPhone. If you want to use the Mews apps and all of these different things, you should be able to use those.

But if you want other apps and other things, then 100%, use them. In the same way, I've got a Salesforce app on my iPhone, but if I want to log in from my browser and go into Salesforce, I'm still fully able to do that. The way that I see it, it's never about closed systems and open systems, because you're just talking about browser windows.

But the way that we ultimately see this is that you should have one system that is able to reveal all the things that are happening in your hotel. From there, you're able to think about how those things happen, but there shouldn't be something else that is in some ways locked off from that. This is the point to me: the openness actually goes two ways.

So there shouldn't be any information in other systems that is locked away from the PMS, either. Which is why the breadth of the APIs that both systems have should be roughly equal. But you should only pay for the things that bring you real value.

So if you're not using our RMS and you're using another RMS, it's important for that interaction to happen. You shouldn't be paying us if you're using a different RMS, but you should be able to use the combined intelligence of those things.

Josiah: That makes a lot of sense. I think 90% of my listeners on this show have iPhones, so I think that analogy makes a lot of sense. One thing that we ran out of time for last time we recorded was agentic AI, and obviously a lot has changed in the world since last December when we recorded. When you think about agentic AI at Mews today, across your products and the potential you see ahead, what does that look like for you and the hoteliers you serve?

Richard Valtr: The main and biggest change, I think, will really be on the distribution side. I don't know how much Meta's Muse and Google will ultimately ask for as their toll on this, but I can't imagine it will be higher than 3%.

Josiah: So big savings after the 20-plus percent many current intermediaries charge. So this could be a big savings.

Richard Valtr: It might not end up being a saving, because it might be that the large OTAs refine their business model. If you think about the economics of an OTA, they charge you 18% to reinvest roughly 10% on things that bring you more business month on month. With that reinvestment, it's a kind of net fee that you can think about of around 8%.

And you could do that yourself. But currently most hotels, whether they like it or not, are making that value judgment and saying, "Actually, I'm okay to pay 8% for that service." And I actually think for a long time that was a good price. It's just that agentic AI upends the necessity to pay that much.

And I think it also gets to a nicer place, to the point about the knowledge base. You should get more guests who really understand what it is that you're selling, and they should make that commitment to coming to you with their eyes wide open.

Sure, on the one side, that creates more pressure on the hotelier. For example, again, in Prague, my family's first hotel was literally on the Old Town Square. The name of the hotel was the Old Town Square Hotel. Now, there were 50 other hotels that were called Old Town Square, or had it in their SEO ranking, and they were not on the Old Town Square.

So that's going to be harder to fake. And you'll now have to say, "Well, my hotel is actually 50 meters away from the Old Town Square, but the Old Town Square doesn't have a lot of great cafes. It has restaurants that slightly rip you off."

So now you need to be able to describe what your immediate neighborhood is and why that's a better fit. That is different, because we have been thinking about it from this SEO perspective: let's slightly tweak it so we look better than we are for these major things. And now we have to really think about all those minor things that really contribute to what type of experience you are actually going to have.

And I think agentic AI will actually help us get to that place, so you're having a much more honest conversation from both sides. And again, I'm a huge optimist on this, but I think that will lead not only to a reduction in your distribution costs, but it should also lead to a better experience, because you have guests who come in wide-eyed about the experience that they're about to have.

And you should also know, to my earlier point, that this guest cares about these things and these things and these things, so you can get ready for that. You can start saying, "Okay, I know that you really cared about these things, so I've gone and organized a tour for you," or suggested what you're going to have.

And again, that puts another bit of pressure on the PMS, because now it's not just distributing the room. It's thinking about what we're doing for the other 24 hours of the experience, because it's now not just the bed that somebody is coming to ask us for. It's also the neighborhood, the local cafes, the local restaurants.

Can I make a booking through you, or have you already made that booking? It's the fact that you have a connection to GetYourGuide, to Uber for transportation, to whatever else. Because now the pressure is on the proprietor to work as that trusted source in the location, which is what hotels traditionally were.

And it's much more about getting you back to the fact that you now become that prism, that filter, that somebody is relying on for those local experiences.

Josiah: I'm excited about that. It's funny, because with all the acceleration and how fast technology is advancing, it's back to the basics.

And if I think back to what we've covered in this conversation, life gets better for associates, the people working in hotels. The economics get better. The guest experience is better. And I love this idea of hotels being platforms for their local city or their community. It feels like all of this is possible.

So I love this world. I'm going to include a link in the show notes where people can reach out to your team. For people listening who reach out to your team, what's something they should ask about that they might not be aware Mews has, or something you're excited about at Mews that they should have a chat with your team about?

Richard Valtr: The one thing that we've released that I think is just the most incredible thing to see live and see working is this feature that we call Automations. If anyone's encountered, for example, Zapier or any of those things, where you can just join up two applications, or two features, functionalities, or two endpoints, or even a wider world of other tech destinations, we now offer this within Mews, and it's really powerful and really exciting.

So if you've ever thought about completely customizing your PMS or your guest journeys, for example: okay, when a guest comes in, I want the GM to get a text message if that guest has been with us a couple of times before. And if the GM doesn't answer, then I want another text message to go out to that guest from the GM, welcoming them to the hotel.

But the GM should get another notification tomorrow morning, when that person is definitely at the breakfast table because they've just ordered. And this is why, again, it's important to have all of these things linked up, so that the system knows they're currently having breakfast, and therefore the GM should come.

Anyway, I've described it in a very, very complicated way, but all of these custom workflows are now possible, and it's just incredible what you can build. One is the team journeys for your GM, or for whoever it is. But the other is thinking about: if the guest makes this decision or has this kind of query, then these are the next actions that we should take.

The way that we're already doing it, this is the programmer part, where I want hoteliers to be thinking that we should be creating these types of algorithms, these if-this-then-thats. This is experience programming, and I'm so excited to think about where that's going to lead.

And it's going to be really, really exciting, because when we still think about programming, we still think in terms of doing it in Claude and doing it in these kinds of ways. Whereas the way that we're thinking about it is: how do we translate the precepts of engineering into the way that you're designing set plays?

I've always likened it to American football, where your team is coming together in a huddle every single morning and saying, "Okay, these are the plays that we're going to play, and this is what we're going to do." And I just think it's going to be a lot of fun for people to be able to design that and talk about it.

And that's the one feature that I'm super, super excited about, and also the future of where that feature goes.

Josiah: That's amazing. I'm going to include a link in the show notes and encourage people to reach out, talk to your team, and get a look at this. It sounds like there are so many possibilities, and you can make sure that this maps to the type of hospitality you want to provide.

So, Richard, thank you for running through all this with me. I always enjoy our conversations, and I learned a ton. Thanks for making the time.

Richard Valtr: Thanks so much.