Agilysys' Frank Pitsikalis: fix the data before you fix the AI

We asked Frank Pitsikalis, SVP & Chief Marketing Officer at Agilysys, to describe the company to someone outside the industry. Agilysys builds the core operational systems hotels run on, the property management system, the point of sale, and the spa, golf and sales-and-catering software around them. Frank didn't start there. He went to Blockbuster. You used to drive to the store, find the tape, take it home, and rewind it before you brought it back, or they charged you. "Be kind, rewind." He says Agilysys is Netflix, and the rest of the industry is still running the Blockbuster store. When those companies add AI, it's a robot that rewinds the tape for you. The point isn't the robot. It's that you're still renting videotapes.

That contrast is the whole pitch. Instead of bolting AI onto old systems, Agilysys built it from the ground up. Underneath sits an AI backbone shared across every application, and the hotel picks how much it wants to spend on understanding a guest: a pricier model from Anthropic for richer insight, a cheaper one where that's all it needs. None of this is new for them. The intelligent guest profile it all runs on came years before the current wave, and Guest Sense AI, the umbrella the features sit under, launched back in 2023.

The data problem comes first

Frank said the same line twice: "The biggest problem in hospitality is data." A hotel knows you at the front desk, then forgets you at the spa. Different system, new profile, and now your history is scattered across the building with no way to tell it's all one guest. Upselling is easy, he can add champagne to your room. Cross-selling is the hard part, moving you from a room booking into a spa treatment or a dinner reservation, because none of those systems talk to each other.

The intelligent guest profile is his answer. One record that follows you across the front desk, the golf pro shop, the spa, the membership office, and pulls the same history up wherever you're standing. On top of that the AI shows what matters right now. He used a guest called Austin, who'd typed a nut allergy into a comment field when he booked. Instead of a clerk scrolling comments to find it, the system reads it, decides it matters, and puts it in front of whoever checks Austin in. "AI is going to take away the manual tasks, the mundane tasks, and allow for true hospitality," he said, so staff can look up from the screen.

Doing AI badly has a cost

On security, his worry is what leaves the building. Agilysys never sends a guest's name or phone number to a language model. It strips the personal details out first, swaps them for a token, and sends only the token to whichever model the hotel chose. The model answers, the system matches the token back to the real guest, and the identifying data never leaves Agilysys.

It's also about shape. A hotel running a dozen separate vendors has a dozen weak doors, and the weak spots aren't the products, they're the joins between them. He called it "MacGyver, bubble gum," all that data passing back and forth across interfaces that were never built to carry it. AI is very good at finding those seams. Run one connected system instead and you've got a single door to lock, and when something does go wrong you patch it once rather than chasing it across every supplier. He brought up Mythos here, the Anthropic model that's been pulled for now, and the fear that a tool like it will let attackers map every weakness in a stack faster than anyone can close them.

Then there's the bill. AI cost balloons quietly, he said, and a hotel that isn't watching can suddenly find it "spent half a million dollars on AI this year" without even knowing it was happening. He said buyers walk the floor and see the demos, voice here, automation there, all of it impressive, and the thing nobody's asking is whether any of it was built to run securely at scale.

The booking question nobody settled

This was the question we put to every vendor this week, what happens to the booking once it moves inside an AI, and Agilysys has gone further than most. In July they'll launch conversational reservations, an AI concierge a guest can call at three in the morning to book a room, a tee time, or dinner, and have it land as a live booking in the system. After that come the MCP servers, the connections that let an outside AI reach into a hotel's live inventory. That's the part that would let someone plan a golf trip inside ChatGPT and have an Agilysys property show up first, because the system can see the rooms and tee times are actually there.

Frank thinks distribution is about to be turned over. AI will democratize search, as he put it, and the hotel's job is to make sure it's there for the direct booking when the moment comes. He hasn't solved the part in the middle. Someone has to build the secure layer between a hotel's inventory and a chatbot, and when we asked who, he said it's still partly on the hotel to manage. And the question we'd carried all week went unanswered here too: when the booking happens inside the AI, who gets paid.

The thing that shouldn't change

We asked what he'd never hand to AI. He didn't have a clean answer, and he said so. He kept coming back to pace, that these tools get shipped too fast, and that the discipline is slowing down enough to be sure before you release. Then he answered it differently. The most rewarding thing in hospitality, he said, is "putting a smile on a guest's face," and that's the part he doesn't want a machine to take. Hotels aren't going to end up all robots at the front desk. The Greek word for hospitality is xenia, and it comes from the word for stranger. In the old stories a stranger at your door might be a god in disguise, so you treated everyone like one. That, he reckons, is the core that shouldn't change, however much of the rest gets automated.