HITEC's Quiet Revolution: AI Stops Demanding Your Attention

I walked into the Indiana Convention Center on Tuesday morning expecting the usual sensory assault — the wall of chrome and LED, the booth babes-replaced-by-product-demos, the giant claw-machine activations, the “ask our AI anything” kiosks that nobody asks anything. HITEC delivered all of it. But by Wednesday afternoon, after my third loop of the floor and a stack of meetings that bled into the lobby bar, I’d written one sentence in my notebook and underlined it twice:

The booths that won this year were the ones I almost walked past.

That is the thesis. HITEC Indianapolis 2025 was a structural inflection — not in attendance, not in dollars, not in announcement volume, but in the shape of what hoteliers actually walked over to talk about. The vendors who built attention machines lost. The vendors who built things that disappear into a housekeeper’s morning, a night auditor’s queue, a GM’s inbox — they won. And the cleanest, hardest-to-argue-with piece of evidence is the E20X Entrepreneur 20X startup competition, where a company called Levee took both the judges’ award and the audience award in a single night.

I want to use this column to unpack why that happened, what the floor felt like, and what it means for the buying cycle running into the back half of 2025.

The headline numbers undersell the room

HFTP’s own wrap puts the show at nearly 6,000 industry stakeholders from more than 60 countries, 360-plus exhibiting companies across 81,500 net square feet of floor. Those are healthy numbers — not record-breaking, but healthier than a couple of the post-COVID years, and the international representation in particular felt up year-over-year. I counted three different Korean delegations between Tuesday morning and Wednesday’s keynote, a sizable Middle East buying contingent shopping property-management replacements, and a real European group up from IBTM and HX adjacencies who’d extended trips to catch HITEC.

But raw attendance is the wrong metric for a year like this one. The right metric is what I started calling “purposeful traffic” — the share of attendees walking the floor with a vendor list, a budget cycle, and a specific replacement decision queued up. That share felt unusually high. I had operator after operator tell me some version of: I’m not here to be sold. I’m here to confirm which of three shortlisted vendors I sign in Q3. The economy of attention in the hall reflected that. Booths optimized for cold-foot-traffic — the ones running game-show buzzers, the ones with mascots, the ones with crowds clustered around a TV — were ghost towns by Tuesday’s last hour. Booths optimized for scheduled meetings ran 30 minutes behind on every demo slot.

This is what a maturing market looks like. The dollars haven’t gotten smaller. The buyer has gotten smarter.

Levee winning both E20X categories is not a coincidence

The E20X is the startup pitch competition that runs as a side stage during HITEC. Six to eight companies, eight-minute pitches, a panel of judges, a live audience vote. In ten-plus years of the format, the judges and the audience routinely split — judges tend to reward technical depth, audiences tend to reward charisma. They picking the same winner is rare. Picking the same winner unanimously is rarer.

Levee took both. The company’s product is a housekeeping operations layer — an AI assistant that sits on top of the property-management system and choreographs the room-attendant workflow in real time. Pulls a guest’s stay status, the priority-clean queue, the inspection backlog, the linen-and-amenity inventory at the floor pantry, and the supervisor’s open tickets. Then it tells a housekeeper what to do next, in their language, on a phone they already carry.

Mark interpretation: that pitch is not a moonshot. There are at least four other vendors in the hall with overlapping decks. What Levee got right — and what I think the room rewarded — is the posture. Their demo did not start with the AI. It started with a fifteen-second clip of a housekeeper, in Spanish, getting a push notification that said in plain language which room was next and why. The AI was implicit. The product was a quieter morning. The judges and the audience were voting for the same thing.

I keep coming back to a line one of the judges used in conversation afterward: We picked the one we’d actually deploy. That, more than any single demo on the main floor, is the HITEC 2025 signal.

”AI that just works” was the only sales motion that worked

Walk the floor Tuesday morning and count how many booth headers contained the word AI. I gave up at forty-something. Count the ones that paired AI with a verb that was also a noun the buyer already knew — “AI room assignments,” “AI guest messaging,” “AI revenue rules” — and the count drops by half. Count the ones where the demo showed AI doing the thing without the operator having to touch a chat box, a prompt window, or a settings page, and you’re down to under a dozen.

That dozen ran the table.

The most useful framing I heard all week came from IDeaS’ Klaus Kohlmayr in a panel recap Phocuswire ran Wednesday, where he argued that the industry has spent two years pitching AI as a feature and now has to pitch it as a substrate. Substrate is the right word. A substrate is something you don’t notice until you remove it. A feature is something you notice every time you use it. Hoteliers in 2025 are exhausted by features. They are buying substrate.

You see this in the smallest decisions vendors made about their booths. The companies who’d gone hard on substrate had taken the AI badges off their booth signage and replaced them with outcome language — “shorter shifts,” “cleaner audits,” “fewer escalations.” The companies still pitching AI-as-feature had bigger AI logos than they’d had in Charlotte the year before, and their booths were emptier by Tuesday afternoon. I am not making a value judgment on the underlying products; some of the loudest AI pitches are connected to real engines. I’m reporting on what worked in the room.

The conversational interface lost its honeymoon

A year ago, if you’d walked HITEC Charlotte 2024, the dominant interaction pattern on the floor was the chat box. Every booth had one. Every demo started by typing a sentence into a text field and watching tokens stream back. That was the visual shorthand for “we have AI.”

The chat box was conspicuously absent this year — or, more accurately, conspicuously demoted. It was still present in most product surfaces, but vendors had stopped centering it. The Canary team, for example, ran a demo built around their staff-facing console rather than the guest-facing chat, and DJ Singh framed it explicitly: guests don’t want to text us, they want their problem solved. Staff don’t want to read a chat log, they want a worklist. That reframing — from chat as the product to chat as one of several substrate inputs — was on every mature booth I visited.

The implication for buyers is real. If you were evaluating vendors twelve months ago and your shortlist was sorted by “best conversational interface,” you are sorting on a feature whose competitive importance has cratered. The new sort key is whether the conversational layer is load-bearing for any task the operator cares about. For most operators I talked to, the honest answer is: not really. The chat is nice. The work is what counts.

Housekeeping and night audit were the floor’s two real categories

Strip away the noise and HITEC 2025 had two emergent product categories — both of them in operations, both of them long ignored, both of them suddenly hot.

The first is housekeeping ops. Levee is the loudest example, but it’s not alone. I counted at least seven vendors with a serious housekeeping pitch, three of them new since last year. The thesis is consistent across the category: the room-attendant workflow has been instrumented poorly for thirty years, the data exhaust from a modern PMS is rich enough to do real choreography, and labor scarcity makes any minute saved per room compoundingly valuable. The math is not subtle. A four-hundred-room property running a fifteen-minute average per clean, with a labor cost in the high twenties an hour, is spending a meaningful share of GOP on this single function. Save ninety seconds a room and the payback period for the software is measured in months.

The second is night audit and back-office workflow. This one is less photogenic, which is part of why I think it’ll move faster than people expect. The traditional night audit — reconciliations, reporting, posting, the close — is one of the last fully manual workflows in a category that has automated everything else. Two vendors at the show had product demos that walked through a fully autonomous nightly close with human-in-the-loop exception handling. I’m not going to name them yet because I want to see whether the demo holds up under property conditions, but the existence of the category was the news. If you’re a regional CFO at a hotel group with twenty properties each running a $45,000-a-year night-audit headcount, you are doing the same math on this as housekeeping.

These two categories were the floor’s center of gravity, and they were operations categories, not guest-facing categories. That is a meaningful shift from the last two HITECs, which were dominated by guest-messaging and revenue-management pitches.

The Four Margins framework I’d lean on, restated for HITEC

I want to gesture at a forthcoming framework piece my colleagues have been working on — what they’re calling the Four Margins of hospitality AI ROI, a forthcoming May framework piece I’d point you to once it’s available: the Four Margins. I won’t restate the whole framework, but the short version is that AI in this industry pays back along four distinct margin levers — labor minutes saved per task, revenue captured per opportunity, retention preserved per friction event, and capital cycle compressed per decision. The reason I bring it up here is that HITEC 2025’s quiet revolution mapped cleanly onto the labor-minutes margin. That’s the lever Levee is pulling. That’s the lever the housekeeping category as a whole is pulling. And it is, not coincidentally, the lever that requires AI to disappear into workflow rather than perform AI-ness back at the user.

The other three margins were represented at the show — revenue-management vendors are still the wealthiest sponsors, retention pitches were everywhere, and you can find capital-cycle stories if you squint at the procurement-tech vendors. But labor minutes is the one with the cleanest 2025 deployment story, and it’s the one the floor rewarded.

Two things to watch through the rest of 2025

I’ll flag two patterns I want to watch carefully between now and the next major North American show.

The first is vendor consolidation in the housekeeping ops category. Seven serious vendors is too many for a category this specific. Within twelve months I expect two or three of them to be acquired by larger property-management or workforce-management platforms — and I’d watch the workforce-management side as much as the PMS side, because the labor data already lives there and the choreography logic is closer to a labor product than a hospitality product. The Yelp-style “ship thirty-five features in two years to lock in the surface area” play that I covered for restaurants in an upcoming May piece — the Yelp AI stack — has an obvious analog on the hospitality side, and the housekeeping ops category is exactly the kind of substrate a larger platform would want to own.

The second is how the next generation of conversational layers gets repositioned. The honest read on the floor this week is that pure conversational AI has lost its momentum as a sale. But the underlying language models are getting cheaper, more reliable, and easier to compose. So the bet for vendors who built their pitch around chat in 2023 and 2024 is not to retreat — it’s to repackage. Expect the next twelve months to bring a wave of “co-pilot” and “agent” product positioning that takes the same model layer and embeds it deeper into operations consoles. The vendors who do this gracefully will keep their seat. The vendors who keep selling the chat box itself will get displaced by substrate-first competitors who can credibly say our AI works without anyone talking to it.

I want to be honest about uncertainty here. It’s entirely possible that twelve months from now the pendulum swings back, that some agent UX cracks the consumer side, and the chat box returns as a real interaction model. The history of this industry’s interface debates is not exactly a one-way street. But I’d bet on the substrate model holding for at least the next two product cycles.

What buyers should do with this signal

I’ll close with three concrete moves for hoteliers reading this with a 2025 budget and a shortlist.

First, re-sort your shortlist by deployment cost, not by demo strength. The best demo at HITEC was not, by any reasonable measure, the best product to live with for three years. The vendors winning on substrate are typically the ones whose sales motion is undersold and whose deployment motion is overbuilt. Ask for the deployment runbook. Ask how many properties they’ve rolled out in the last six months. Ask the customer references about the implementation, not the steady state.

Second, bias the operations categories over the guest-facing categories for this budget cycle. Guest-facing AI is not done — there are real products in revenue management, in personalization, in messaging — but the labor-minutes margin is the one with the cleanest payback math right now, and operations is where that margin lives. If your CFO is asking for ROI documentation, operations vendors will give you the cleaner case to write up.

Third, do not over-index on the E20X winner. Levee earned the win and I think they’re a real company. But the audience and the judges were voting on a category, and there are credible alternatives. Talk to two or three of them before signing. The point of paying attention to E20X is the category signal, not the individual logo.

Coda: the booth I almost walked past

I’ll end with the booth I keep thinking about. It was a small one — ten by ten, two product people in polos, a single monitor running a recorded demo on a loop. No mascot. No giveaway. No “AI” anywhere in the signage. The product was a maintenance-engineering coordination tool — the kind of thing you’d dismiss in three seconds if you were scanning for novelty.

I almost walked past. What stopped me was the loop on the monitor: a maintenance ticket coming in from a guest-side messaging system, the AI parsing it, dispatching to the right engineer’s queue with the right priority, and closing the loop with the guest in their language once the engineer marked the ticket done. The whole flow was thirty seconds. No human typed anything except the original guest message and the engineer’s two-word resolution note.

That booth wasn’t winning E20X. It probably isn’t getting written up by the wires this week. But it was doing the thing the rest of the floor was trying to talk about — letting the AI carry the choreography and getting out of the way. I left thinking that in twelve months, that booth has a product that an operator either deploys without thinking or evaluates against three competitors who all look exactly like it.

That is what a quiet revolution looks like. Not the booths with the lights. The booths with the loops.

HITEC heads to San Antonio June 15-18, 2026. Between now and then, I’d watch the operations categories closely, watch the workforce-management M&A column for housekeeping plays, and watch the language vendors pivot away from chat-first positioning. The thesis isn’t that AI loses its centrality — it’s the opposite. The thesis is that AI gets quiet enough to be everywhere, and the loudest vendors are the ones who didn’t get the memo.

I’ll be back on the floor in San Antonio with the same notebook and, I hope, a similarly short underlined sentence.

— Sofia leads Vibe Check vendor reviews for TableTransfers. Tips: vendors@tabletransfers.com.