Your Network, Not AI, Is the Real Hotel Problem

I just got back from HITEC in San Antonio, and there was a lot of conversation about AI in hospitality. What struck me was the disconnect. There's one group of people who think AI is going to replace customer service or customer satisfaction, and then there are the well-versed operators who are looking at the ways this could actually help them, and the gap between those two conversations is enormous.

The operators aren't talking about replacing anybody. They're looking at the back office pain, the booking pain, the rewards and the things they give different clients who stay at their properties. That's where the value is, it's nowhere near the front desk, and almost nobody in either group was talking about the part that decides whether any of it actually works.

What has to be true before AI works at a property?

They need a good, solid technology foundation in order to run any of this stuff. That's really the key, and their network has to be ready to accept it. If they're going to start using AI to run back office functions, or to communicate from point of sale, or from guest management, they need to make sure they have a solid IT foundation, or they simply won't be able to take advantage of it.

That covers everything from the wiring being updated, to the equipment that routes everything, to the Wi-Fi they use on their devices, to the way people connect to that Wi-Fi, all the way through to security, which ends up being pretty key in all of this. So I wanted a second opinion from somebody who runs properties rather than designs networks for them, and I asked Hershal Patel, a hotel operator, whether that matched what he sees from the inside. He agreed, and then he immediately reordered my list.

"I think the hotel needs to know what they want, first of all," he told me. "And then, of course, they have to have good data infrastructure to support an assistant like that without lag."

That's a better starting point than mine and I'll take the correction. I come at this from the network side, so my instinct is to audit first and ask what it's for second. He's right that an operator who hasn't decided what problem they're solving will buy infrastructure for a use case that never arrives.

Where do operators actually get exposed?

Once you know what you want, I'd want a security audit first and foremost, before automating a single function. I'd want to make sure that whatever they're doing internally, their systems are set up for it, because we've seen some pretty creative hacks in the way people are taking advantage of AI systems.

If they tie an AI system into multiple agents, then those agents now have access to their email, into their bookkeeping, into all their accounting software, and whatever else is connected. I'd want to make sure their network is impermeable from the beginning, and that the pieces that don't need to see each other are segmented so they can't.

What makes independent operators specifically vulnerable is that a lot of them buy off-the-shelf equipment built for home and residential use, or they don't really understand the implications of that technology, so they install gear that goes unpatched and unnoticed and isn't capable of supporting the firewall rules they think they have.

If somebody wants to attack something, they can look at those consumer-based devices, see what's connected online, and attack that network precisely because they know it's a residential device that's supposed to be providing a security posture and isn't. Smaller operators don't make big investments with a security-posture-first mentality, which leaves them open to a number of different attacks, and not only from bad actors: we're now seeing it from AI agents as well, which can identify where AI is being used and go after those properties knowing they can get through the lapses.

The power of AI in hospitality is going to be in appropriately directing agents to assist in service recovery and curated guest experiences. — Hershal Patel, hotel operator

Does AI replace hospitality?

Here Hershal was sharper than the whole conference, and he's the one who has to live with the answer.

"As a guest-facing tool, there really shouldn't be a fear, because as we move forward I think the human element will always have to be there for true hospitality. I don't think anyone likes robots being the only interaction they have. What AI needs to do is augment that by creating a near-perfect employee or agent that will register and respond to guest issues appropriately, and convey information to the people or staff who are empowered to do something about it, and recover the guest experience before it's too late."

Service recovery. That's the use case, and it's a completely different thing from the chatbot everybody is arguing about. A guest has a bad experience, somebody notices it in the moment, and the right person is told while there's still time to fix it. None of that requires replacing a human being, and all of it requires systems that can actually talk to each other.

What breaks when the systems don't talk?

With the speed AI is moving at right now, it isn't just a software tool. It's the way things are constantly being reinvented, the way APIs are being opened up, the way MCP services are being used to communicate from system to system. So it isn't just learning the nomenclature or the jargon of the hotel industry. It's about understanding the way the systems function within each other and communicate with each other.

If you don't have guest management talking to bookkeeping, talking to point of sale, you lose transactional information from individual guests. You can't see who's eating at the steakhouse. You can't see what the expected revenue at the restaurant is based on hotel occupancy. You lose visibility across all of those silos, and every one of those blind spots is a decision somebody is now making on a guess.

What does the load actually look like?

Hershal gave me the best example in either conversation, and it's the kind of thing you only learn by operating a property.

"We've been prompted to increase bandwidth to our larger hotels at least twice during snowbird season in the winter months on South Padre Island," he said. "What we did not anticipate is that winter Texans love to use their devices, and they check in multiple devices, including a streaming stick for the TV, an iPad, a couple of phones, and maybe even a fifth device, all connected at the same time, usually using at least one or two of them simultaneously, and across the whole hotel because the weather isn't conducive to being outside all day."

Read that as a design brief rather than an anecdote. Every room is five associated devices instead of two, most running at once, concentrated indoors across the whole property for weeks, and they had to go back for more bandwidth twice in one season before they caught up. His fix since then has been to run what he calls "way, way overkill" on bandwidth and refresh the hardware inside roughly every five years. Nobody in that story was doing anything exotic either. They were FaceTiming their grandkids and streaming old TV shows, which means the load that broke the assumption was ordinary guests behaving normally, and that is exactly the load an AI rollout lands on top of.

Why do hotels get Wi-Fi wrong?

You find a lot of hotels that think about Wi-Fi purely from a coverage perspective. They just want to make sure it permeates the walls, and they don't think about handoff. They don't think about the way Wi-Fi works when you're using voice calling, or when you need real-time telemetry, or when you need the ability to track a device, or when you want to offer a great guest experience by doing something like Wi-Fi offload or cellular offload for calling.

Independent operators generally don't put the emphasis on technology that the larger brands do unless they have a substantial budget for it, so we need to make technology smarter, so it can overcome the humps those operators face from a budgetary perspective. On which point, I asked Hershal what this actually costs, expecting the answer that usually kills these conversations. He said: "In relative terms, not much at all. Your data bill every month could be as little as a few hundred bucks, or as much as a few thousand, so it depends on the size of the operation and what the needs are." A few hundred to a few thousand a month is not the barrier people think it is, and it is dramatically less than one bad season of guest reviews.

Do this

• Decide what you actually want AI to do before you buy anything, because Hershal is right that infrastructure bought without a use case is infrastructure bought twice.
• Run a security audit first, especially if anything you're considering will connect an agent to email, bookkeeping, or accounting.
• Then run a network audit, sized to the services you actually intend to automate rather than to the building.
• Then run a Wi-Fi audit and check coverage in every corner of the property, using a heat map rather than a walk-around and a guess.
• Go look at the router at your property right now. If it's a residential box, or it's off-the-shelf gear you haven't updated in a few years, that is the finding and you already have your first project.
• Get on a reliable pipeline. Hershal on what that means today: "business grade fiber. Symmetrical fiber connection." Then evaluate your signal coverage inside the property, using that backhaul to its fullest potential and making sure every corner is covered.
• Map whether guest management, bookkeeping, and point of sale can actually see each other, because that plumbing is what any useful AI will run on.
• Size for five devices a room, not two, and size for them running simultaneously indoors during your busiest weather.

One of the great things about AI is that it can help bridge that gap, so that technology helps make these properties better instead of technology being the thing that hurts them the most. But that only happens on top of a network that was built to carry it, and right now most independent properties have not built one.

FAQ

Will AI replace hotel staff?
Not for guest-facing hospitality. Both of us landed in the same place on this: the human element has to be there, and the real value is AI directing the right issue to the right empowered person fast enough to recover the guest experience.

What should an independent operator do first?
Decide what you want it to do, then run a security audit, then a network audit, then a Wi-Fi audit. In that order, and before any purchase.

What does the network side of this cost?
For connectivity, an operator running multiple properties put it at a few hundred to a few thousand dollars a month depending on the size of the operation, plus a hardware refresh roughly every five years.

Why is residential equipment such a problem?
It goes unpatched, it can't support the firewall rules operators assume they have, and it's identifiable from outside. Attackers specifically look for it, because a consumer device advertising itself on a business network is a reliable sign the security posture behind it isn't real.

Want this kind of thing weekly, in audio? That's the show.