The Big PMS Debate: a closed-door discussion between the world's leading PMS CEOs and hotel CIOs

It does not happen often that the people who run hotel technology and the people who build it sit in one room and talk openly. They usually meet across a sales table. This session was made to take that away. It ran on the sidelines of HITEC in San Antonio, which is part of why it was possible at all. It brought together operators and suppliers who are based across different regions of the world and are rarely in the same place at the same time. For ninety minutes, CIOs and CTOs from some of the largest international hotel groups sat with the CEOs and founders of some of the most widely used PMS suppliers. No stage, no slides, no sales pitches, and no audience. Among the hotel groups taking part were Radisson, citizenM, Mandarin Oriental, Wyndham, Aimbridge and Choice. The PMS suppliers in the room included Apaleo, HotelKey, Shiji, Oracle, Infor and Mews. One participant joined in remotely. Everyone agreed to one condition: nothing said in the room would be tied to a name afterward.

The session ran under the Chatham House Rule, which means we can report what was discussed, but not who said it. We did this on purpose, so that all participants could speak freely, knowing that nothing they said could be held against them afterward. That is why the views here belong to "a CTO" or "one supplier," never to a name, and everyone spoke more openly because of it.

The session was hosted and moderated by Hospitality Net's Floor Bleeker. We used live polling during the discussion, so alongside the talk we also have a set of votes from the room. The polls are shown without names too. They are counts, not quotes.

What AI actually changes

Floor started by asking each participant for the one biggest way AI will change the industry. There was more agreement than you might expect, which is a good sign for a first session: labour and cost was mentioned multiple times. The idea was not only that the same work needs fewer people, but that it changes what one role is responsible for. Instead of a front desk person and a separate bell person, you get one employee who handles many tasks across the hotel.

Another topic was the cost of distribution. Some expected it to fall a long way, maybe close to free. The reason is that the commissions hotels pay to online travel agents and other booking channels are seen as unsustainable at today's levels, and as AI changes the way guests search and book, those fees are expected to drop.

Personalization came up as well. It gets much easier, and guests are willing to pay for it, from a room set up the way they like to being greeted by name. Luxury is well placed for it, because the service model to deliver it is already there. And other segments can use the same tools to move upmarket, since technology means you no longer need as many people to make that kind of service work.

One idea came up on both sides of the table, from buyers and suppliers alike: that everyone is becoming a builder. Instead of buying every piece of technology, teams can now build their own tools with AI and low-code, shaped the way they want them to work. That changes the job of the technology leader, from buying software to enabling the people who build it.

One person added a caution. Hotels employ very large numbers of people, and that work is about to change a great deal. The industry carries a responsibility to those workers, to help them into new roles as the old ones change, not only to count the savings.

Everyone agreed that AI matters. Where they divided was on what it is mainly for. Some saw it first as a way to cut cost and labour. Others saw it first as a way to improve the guest experience. That same divide came back in almost every block that followed.

Wins and mistakes of the last decade

Floor asked the group to name one thing the industry got right in the last ten years. The votes grouped around three things: the move to cloud, open systems and open APIs, and keeping people and the guest at the centre.

Most of the discussion stayed on that last one. The point was that travel is personal in a way few other things are. A stay is not only a bed for the night. It is your comfort, your mood, and everything around it, the food, the wellness, the service. Because so much of it depends on people dealing with people, human contact is still what hospitality is built on, and the room felt the industry had been right to protect that. Suppliers said they try to keep the guest in mind, even though the hotel, not the guest, is their direct customer.

But the same discussion exposed a gap the industry has not closed. For twenty years hotels have talked about using guest data, and they now collect an enormous amount of it, yet very little reaches the guest at the moment of the stay. One example came up more than once: nothing in a normal booking captures whether you are travelling for business or for leisure, even though that single fact would change almost everything a hotel should offer you. Someone arriving late for a conference wants to get to sleep. The same person at a resort with their family wants to hear about everything on offer. The data to tell those two apart is rarely asked for, and rarely used.

Where it is done well, it shows. One operator described using climate control that learns a guest's preference, so someone who always cools their room down arrives to find it already set that way, which keeps the guest happy and lowers the energy bill at the same time. Another said it will go a long way to fix, at a guest's next stay, a problem they had at a different property. What the room wanted was to make this kind of data easier to act on at the property, and faster.

Then Floor asked the opposite: what is one thing the industry got wrong? The votes grouped around payments, building on legacy and all-in-one thinking, and integrations. In the discussion, the funding model was the one people kept coming back to. The problem is that the money to pay for technology is split and hard to move. It comes from several departments at once, marketing, reservations and others, each with its own priorities and its own limited budget. Because no single budget owns it, it is hard to fund one large change that works for everyone, rather than a compromise that half satisfies each department.

On top of that, a hotel-group technology leader cannot simply decide to spend. They often have to convince many separate owners and franchisees to pay for anything, one by one. As one put it, "the hardest job in this industry is the one where you have to convince a majority of your franchisees before you can do anything at all." That makes it slow to move money out of day-to-day operations and into technology, even when the case is good. In industries where the value is clearer and fewer people hold the decision, the same change is far easier to fund.

There was sympathy for the vendors too. A vendor signs a deal and then has to make it work across thousands of properties and chase payment from each, without being able to switch anyone off, because that would upset the brand.

One suggestion was to change how technology is paid for. Instead of burying the cost across departments, treat it as its own direct line, and then track over time how that spend pays back. It would make the value easier to see, and the funding easier to defend.

Is hotel tech really behind airlines and retail?

The industry is often told that hotel technology is ten to fifteen years behind airlines, retail and banking. Floor asked whether that is true. Seven said true, two said false, and three picked a third option: it is a self-serving narrative the industry keeps repeating.

We spent more time on this question than any other. The case for "behind" was simple: you can walk onto a plane with your phone, but getting into a hotel room still takes several steps. The pushback picked apart the comparison itself. One supplier said the link to retail and aviation gets made because it sounds logical, but that it does not hold up. Retail is a much simpler flow, often one transaction and one loyalty program. Hospitality has many departments, many systems and many rules, and a guest is with you for many hours across all of them. Building something that works across all of that is a different order of problem.

Airlines were described the same way, as a more concentrated business with fewer decision-makers, so vendors can focus and the industry can agree on one approach. A hotel, by contrast, looks after a guest across many touchpoints: rooms, food, spa, shops, events. Each department often runs its own technology, sometimes with several systems to choose from. And the rules change from country to country in a way that buying a coffee does not. One example given was checking a guest in: a coffee is bought the same way in the United States and the Netherlands, but checking someone in can mean collecting a passport and reporting to local authorities, so there is no single global process to build for.

A common view was that the technology mostly exists, and that the real problem is adoption. Change has to be sold to many owners and franchisees, not ordered from the top. One participant argued the "behind" line is a story the industry tells itself, and that given the complexity and the fragmentation, hotel technology is not as far behind as it is said to be.

The next poll asked who actually leads innovation in hotel tech. Hotel groups got six votes, startups three, and established PMS vendors none, in a room that was half made up of vendors.

Nobody took that to mean vendors do nothing. Innovation is a joint effort, several said, and it runs both ways. A brand brings a problem and a vendor offers to solve it, or a vendor brings an idea and asks who wants to try it. Startups often prove a new idea first, and once it works, the big PMS platforms build it in so everyone gets it. There was a counter to this too: startups on their own often bring ideas that are half formed, and the strongest results come when hoteliers and vendors design something together rather than either side alone.

But two worries stood out. The first is that when only the big brands and the big vendors set the direction, the industry keeps circling the same ideas, because most of them came up through the same world, and the smaller a company is, the less say it has. The second is access for startups. A small company with good technology is often seen as too small for a big brand to work with, and closed systems on both the brand and the PMS side make it hard and expensive to connect at all, so promising ideas never get through the door.

One more question came up here, connected to the earlier point about everyone becoming a builder. If startups and brands are where the new ideas come from, and vendors mostly integrate them, then the PMS vendors risk becoming the layer that others build on top of, rather than a source of new ideas themselves. No one in the room had a real answer to that.

We say guest experience, we fund cost saving

The next question was which AI buzzword is most overrated right now. This one was a quick word cloud rather than a debate, but the result is worth noting: "agentic" came out well on top, ahead of lighter entries like "AI slop," "tokenomics" and "saaspocalypse." The term the industry talks about most right now is also the one this room found most overhyped.

Then Floor asked where the AI focus really lies today. On a scale where higher means more focus, the room's averages ranked like this:

• Cost saving on labour, 4.4

• Tools for employees, 3.7

• Guest experience, 3.6

• Rates and revenue, 3.4

• Other cost saving, 3.2

• Distribution, 2.8

Floor put the contradiction to the room directly: earlier, people had said guest experience is what sets the industry apart, yet the strategy votes lean to cost and labour. The reason came back to who pays. Hotel owners fund the technology, and owners look at the balance sheet, so a project has to show a saving before they will back it. Even when the real value of an AI project is a better guest experience or more revenue, it gets presented as cost saving, because that is what an owner will approve. And many owners arrive already sure that AI will cut their costs, without a clear idea of how. As one person put it, "every owner walks in certain that AI will save them money. They just cannot tell you how." That sets a hard expectation to meet.

One supplier pushed against the labour-saving reflex. The argument was that aiming AI straight at cutting labour is the wrong starting point.

If cost saving is your starting point, you will get AI wrong.

Better to point it at a real problem, like a better guest experience, and let the cost saving fall out as a side effect rather than chase it head on. Another, more careful approach was also described: run workshops across departments, collect hundreds of ideas, then cut them down to the three or four areas where AI brings real value. Two came out of that as concrete: AI in the contact centre, to support the agents and improve the guest relationship, and coding agents to speed up building new features.

This linked back to who pays. If personalization is where AI adds the most value, someone has to fund it, and the room split on who. One view was that personalization is the brand's job, because the brand owns the guest and charges the owner a fee partly for that, while the owner should carry the cost of labour and operations. That raised a pointed question: owners pay brand fees partly for the guest relationship, so are the brands actually delivering the personalization those fees are meant to buy?

What CIOs want, what CEOs build

Floor then turned to the question the session was built around: in five years, will the PMS be more or less central to the tech stack? Seven said less, five said more.

The split had real thinking behind it. The people who said less central argued that the PMS will not look like today's PMS at all. It will either take on far more, or break into several connected systems of record, one built around guest data and one around the day-to-day operation, linked intelligently instead of squeezed into a single core. Part of why the CIOs and CTOs leaned this way is that many hotel groups have gone asset light. Their attention has shifted to distribution and to the guest relationship, and away from running each individual property, which is exactly what a PMS does at its core.

The people who said more central made the opposite case. As AI and self-built tools spread, you still need a stable, high-performance system of record underneath everything, something that holds the transactions reliably while the activity happens in many other places. In that view the PMS becomes the dependable core, even as more of the experience moves outside it.

One point came back more than any other. The PMS has always been built around the room, and several argued it should be built around the guest instead. A hotel is not only its bedrooms. It is the restaurant, the bar, the lobby and the spa, and a guest moves through all of them. A system that only knows the room misses most of the stay.

Not everyone agreed the guest system and the operational system should merge into one. The case against was that a record sitting above the property, like loyalty or a full guest profile, does a different job from the engine that runs the property day to day, and that forcing them together makes both worse. Others pushed back that personalization cannot stop at check-in and check-out, and that the restaurant, the spa and the gift shop are exactly where guest data is needed, and exactly where the PMS usually is not.

Underneath the theory sat a very practical problem: too many systems, and too many logins. Staff have to move from one screen to another, and things fall through the cracks. One supplier gave a clear example. After a single separate login was removed, so staff no longer had to remember to open another system to offer an upsell, upsells rose sharply. The tool had not changed, only the friction around it.

To test one part of this, Floor split the room in two, the hotel-group CIOs and CTOs on one side and the vendor CEOs on the other, and asked each side the same question from its own angle.

The CIOs and CTOs were asked whether their PMS vendor offers truly open architecture at no cost to them: four said yes, three said no. The vendor CEOs were asked whether they let customers integrate what they want at no cost to them: four said yes, one said no. Put side by side, the two answers do not match, and that gap set up what came next.

The vendor CEOs saw themselves as fairly open. The CIOs and CTOs did not quite see it the same way, and the discussion explained why. Both sides agreed the purely technical barrier has mostly fallen. Building an integration is not the hard part anymore. The hard part is commercial. Someone has to fund the per-transaction fees, and there is often a stack of parties between the hotel and the startup, each taking a cut, which is where a promising idea quietly dies. A newer worry is the cost of scale. As AI agents start making huge numbers of calls, vendors have to pay for the infrastructure to serve them, so "integrate for free" runs into a real bill.

There was also a fairness problem for startups. A small company usually cannot afford a large upfront fee just to try something, so the door is closed before the idea is even tested. A few directions came up. One is a shared integration standard, something like MCP, where everyone agrees on the pattern and lets agents handle the detailed work, which would open access far wider. Another came from a vendor who described building tools to help startups integrate, and then stepping out of the middle, leaving hotels to choose who they work with.

Last, Floor asked who owns the customer data, the profiles, the transactions and the loyalty records. Five said the brand, three the hotel owner, three something else. Nobody said the vendor. After that, the simple answer fell apart. One view was that the guest owns their own data and can take it with them, while the hotel owner holds the dataset. But the real cases get tangled fast. Who owns the record of a keycard opening a door, or what a guest orders from the bar? It could be the owner, the brand that runs the system, or the guest. Distribution agreements settle it in some cases, and a management company can lay claim to it in others. People kept returning to one rule.

Whoever signs the contract owns the data. Everything after that is an argument.

And today, an AI is trained only on a single property's data, not pooled across a brand, unless a specific agreement allows it. That makes it slow to learn, compared with, say, a car maker that trains on data from millions of vehicles at once.

The last big question was whether the industry should share more data and train models across many hotels, not just one. The answer was consistent. Personal data stays protected, and no one wanted to give that up. But there was clear interest in sharing anonymized, aggregated patterns, the kind that help everyone without exposing any single guest. The example given was upsell ordering. If the industry learns across thousands of properties that early check-in is the offer most guests accept, every hotel could show that option first by default and sell more, with no personal data changing hands. The room drew the line there: share the patterns, protect the person.

First of its kind

Floor Bleeker hosted the series for Hospitality Net and moderated the session. Looking back on it, in his own words:

This was the first time a room like this had come together, and it delivered more than we could fit in the time we had. I had prepared far more questions than we got through, maybe four times as many. That was not a problem. Everyone was involved, everyone had a view, and every question opened into a real discussion. We could have moved faster and covered more, but it would have been a poorer session for it. I would rather run long because people are talking than finish on time because nobody is, and the fact that we ran out of time is one more reason to do this again.

The thing I keep coming back to is the room itself. This group had never sat together before. Some of the largest hotel groups and the main PMS vendors in one place, and many of them had not even met each other. That alone made it worth doing.

For me the clearest finding was one difference between the two sides. The hotel groups are looking to optimize cost, while the PMS vendors are building for guest experience. Once you see that gap, a lot of the rest of the discussion makes sense.

The other thing I take away is that innovation is being held back from both directions at once. Vendors make it hard and expensive to integrate, and hotel groups do not open the door to startups easily. And on data, my honest read is that nobody really knows who owns the customer data, even though it is one of the most important pieces for anything we want to do with AI.

None of these are technical problems. They are about money, about who can integrate, and about who owns the data. That is the part that stayed with me.

We covered a corner of what we set out to cover, and that is exactly why there should be a next one. Everyone in the room said they wanted to do this again, and so do I. Next time with more time, and probably more than once.