HITEC 2026 so far: where AI feels thrilling and unsettling at once

By the end of the first day at HITEC I had two feelings running at once. I use AI every day and it makes my work better and faster, so I walked in excited about what is coming. I also walked out of the sessions a little uneasy, because the more sure everyone was that AI will reshape hospitality, the clearer it became that no one can tell you where it actually lands. That mix is the most interesting thing about this moment, and it ran through everything we saw. Here is what stood out.

Day one: the ideas

The session that stayed with me was "Distribution 2.0," and the idea at its center came from Fredrik Sjoberg of the agentic AI company Maison. He calls it the return of the travel agent. Decades ago most travel was booked through human agents who took the friction out of it; by 2030, he argued, most bookings will be made by agents again, only artificial ones. The bright side is more travel and a real opening for independents, since a unique hotel that was invisible in the old filters can be found by an agent working a global audience. The worry is the toll. Human agents took around 10 percent, OTAs take 15 to 25 percent, and the App Store held 30 percent for years simply because it could. Sjoberg would not promise distribution costs fall, and his advice to anyone walking the floor was blunt: make your content machine readable down to the exact pet policy, accept that rate and content parity are over, and ask every vendor whether there is an MCP endpoint you and your guests can connect to.

The afternoon session, "The Dark Side of AI," was built to argue the downside, and it did. AI is amazing and the train has left the station, the panel agreed, before quoting Max Tegmark: the view gets better and better right before you drive off the cliff. The energy bill was sobering, with AI consumption roughly tripling in a year and most of that power still coming from coal and gas. Carl Winston of San Diego State, whose program certifies sustainable resorts, admitted he can no longer define what a sustainable resort even is once AI is in the mix. There was the cognitive cost too, a study where the people who let AI do the writing could not remember what they had just produced, set against the reminder that writing itself was once feared as the ruin of memory. The close was a two-sided ledger, energy use against faster innovation, deepfakes against AI defense, and a plea to steer this thing together while we still can. That is almost exactly where I landed myself.

The day's headliner put the operators on stage, and the question that mattered was whether brands survive AI at all. Lennert de Jong of Another Star and Scott Strickland of Wyndham both argued loyalty holds: AI just becomes another channel, and soon you complete the booking inside the assistant and still earn your points. Floor Bleeker disagreed. If agents start talking to agents, he said, brands lose their grip on distribution and have to win owners back through guest experience instead of loyalty.

The sharpest moment was over something simpler, whether hotels still need people at all. Bleeker argued that staff are often a barrier between the guest and good service. Keryn McNamara of Aimbridge pushed back hard: this is the business of humans, she said, and she does not want to see a fully autonomous hotel in her lifetime. Nobody on stage settled it, which felt right.

Day two: onto the floor

The Tuesday headliner, James Taylor, pulled the mood back toward optimism, with a talk on pairing human curiosity with AI rather than being replaced by it. His most useful idea was the competency penalty: people hold back from AI not because it is hard, but because they fear that using it makes them look less capable, and the research he cited found women and older workers feel that pressure most. Name it out loud, he argued, and adoption follows. After a day of warnings, it was a good reminder that the human worry about AI is often social, not technical.

Two things of our own this week. We recorded our first Boardroom Reboot podcast, with Floor Bleeker hosting Richard Valtr of Mews, which turned into less of a product chat and more the story of how Richard got into hospitality and built what he built. It will be on Hospitality Net soon. And the Hotel Yearbook "AI Everywhere" launched here, the technology edition given over entirely to AI, with 40 expert articles, 26 solution snapshots and a hospitality AI glossary.

Then the floor opened, and we worked through a run of booth conversations. We are running these differently this year. Instead of a commercial video, we record a real conversation and write it up as an article, because that is what people want to read. What follows is the one thing we took from each, not the whole of it.

On the operations side

Agilysys put the real problem plainly, and it is not AI, it is data. Hospitality is so siloed that a guest becomes a new profile in every system, and personalization only works once that data is joined up. It showed how: a guest types a nut allergy into a comment field at booking, and the profile surfaces it for the front desk on its own, then carries it from check-in to the spa to the golf shop. Frank Pitsikalis put the lesson well, that clearing the busywork is what lets staff be more human, not less.

When the storm grounded flights into San Antonio, Infor's team built an agent overnight that fed live delays into the housekeeping board, so a room is not turned for a guest stuck on a plane while early arrivals get readied first. David Poprawka's bigger point was that the competition has changed: hotels can now build their own agents, which puts vendors up against Google, OpenAI and Anthropic as much as against each other.

Know what to automate, and what to leave to the experts. That was Trybe's discipline. Its new overnight engine sells an inclusive spa break and splits it automatically across the spa, the PMS and the restaurant system, and Ricky Daniels says three pilot properties ran more than £150,000 of these bookings in about two months. But on autonomous revenue management he drew a firm line, that some things you cannot just vibe-code.

Actabl's answer to hallucination is technical rather than flashy. Instead of handing a model a spreadsheet and hoping, they trained it on how the data relates and taught it to write the query, so the numbers hold up. Jerimi Ford's description of the alternative, arguing with the AI for a while before you get what you want, will sound familiar to anyone who has tried. The sharpest detail was where they stop: AI reads a boiler's nameplate to log it, but does not diagnose faults, because a wrong call could be expensive or dangerous.

On distribution and AI search

Start with a number: about 30 percent of calls to a hotel go unanswered, many from people trying to book direct. Canary's voice AI answers every one, and at worst leaves the manager a list to call back. Its rule for rolling this out is simple and sane, analyze, then assist, then do, with the AI only acting once the hotel switches it on. SJ Sawhney was candid that nobody, the builders included, fully understands the line between AI and agentic yet.

Why do hoteliers still override their revenue systems? Because the old tools are black boxes, Lighthouse says, so people do not trust them. Their fix is an assistant, Ernest, that explains itself: you can ask why it moved a rate, and tell it in plain words never to move more than a set amount without asking. Brett Kohn framed it as a teammate that can analyze, recommend and act within guardrails, the example being a group displacement analysis that used to mean a manager and a spreadsheet.

The most useful thing Cendyn said was what is not real yet: nobody is booking a hotel inside ChatGPT today, people read the answer and leave to book, and an app you install in a chatbot is not the future. Their bet is Google as the channel where the booking can really happen. Their new tool checks what the assistants say about a hotel, scores it against rivals and flags where the website disagrees. Michael Bennett's hopeful note matched Sjoberg's: with most European hotels small independents who cannot outspend Booking.com, authentic content is now worth more to a model than SEO ever was, and there is a real chance to win share back if the industry moves first.

If one thing ran through all of it, it is what the sessions kept circling: the bottleneck is clean, connected data, not the model. And the question nobody answered, on stage or at a booth, is who collects the toll when the booking happens inside an AI.

Where this leaves me

So that is the first two days. I came in excited and I still am. The possibilities are real, and a lot of the work on the floor is genuinely useful, most of it quietly, in the back office, where the wins are concrete and far less contested than the distribution drama on stage. I also leave the sessions a little uneasy, and I think that is the honest place to be. The best line of the week was not a vendor's. It was the reminder that this is either going to be great or it is not, so we had better work together to make it great. Where it actually goes, the future will tell.