Inside Marriott's 1,000-Hotel Tech Cutover (and What's Different from Hilton's Approach)

It is 6:42 a.m. on a Thursday and I am standing behind the front desk of a Marriott-flagged select-service hotel off an interstate exit I will not name, watching a night auditor close out a shift on a system she has been using for eleven days. She is fast. She is also annoyed, in the specific way operators get annoyed when something works but not the way their muscles remember. “It’s fine,” she tells me, without looking up. “It’s just not the thing I learned on.”

That is the sentence I came here for. Marriott CEO Anthony Capuano said it differently on the May 6 earnings call, but he was describing the same moment, at scale: “Just yesterday, we transitioned our 1,000th hotel over to our new tech ecosystem.” A thousand front desks, in other words, where some version of that night auditor is fast and annoyed. The cutover is the story. The strategic frame around the cutover — why Marriott is building with vendors instead of standing on top of them the way Hilton is — is the story under the story, and it is the one the call did not slow down for.

I covered the Hilton side of this two weeks ago, after Chris Nassetta walked through his AI Planner deployment on the April 28 call. Sofia ran the product-level Vibe Check on Marriott’s stack last Wednesday, the day Capuano’s quote dropped. This piece is the operator case study sitting between those two: what Marriott actually shipped, what it actually replaces, how owners are paying for it, and why the fee economics that fall out the back end will look different from Hilton’s even if both companies are right about the technology.

The headline Capuano didn’t lead with

If you read the Marriott Q1 release straight through — Q1 RevPAR up 4.2%, adjusted EBITDA at $1.4 billion (a 15% lift), adjusted EPS at $2.72 (up 17%) — you would not necessarily catch that the company quietly crossed a threshold it has been working toward since the back half of 2024. The press release leads with financials, as press releases do. The 1,000-hotel number lived in the prepared remarks, in a sentence Capuano delivered in the matter-of-fact tone he uses when something has, in fact, gone well.

“Just yesterday, we transitioned our 1,000th hotel over to our new tech ecosystem.”

That is a load-bearing sentence. Marriott runs roughly 9,400 hotels worldwide. A thousand is not the whole portfolio. It is, however, the proof that the playbook works property-by-property, brand-by-brand, region-by-region — which is the part of any global cutover that decides whether the next nine thousand are a sprint or a slog. When you talk to owners who have been through it, they describe the cadence as “weeks not days, but not months either.” A franchisee I had coffee with on Friday morning, a multi-property operator in a secondary market, put it this way: “It was the second-worst weekend of the year. The worst was the one we expected it to be.”

The 30–35% of Marriott’s $1.05–$1.15 billion 2026 investment spend earmarked for digital and technology transformation is what funds this. That is roughly $315M–$400M of corporate capital aimed at the stack, give or take, on top of whatever owner spend the cutover triggers at property level. Numbers of that size at a fee-based asset-light company are not abstract. They are a decision to build something the firm will charge for later.

What the new ecosystem actually replaces

Walk into a Marriott-flagged hotel pre-cutover and the back office is a museum. The PMS is something the night auditor has been using since before her current manager started. The CRS is something else. The revenue management surface lives in a third tab. Loyalty interactions route through a fourth, with a fifth window open to the brand portal for whatever today’s compliance ask happens to be. Reporting is a Frankenstein of exports.

Walk into the same property after cutover and the topology compresses. Not to one window — nobody is claiming that — but to fewer surfaces, with the data layer underneath them stitched together in a way that lets the loyalty record, the rate plan, the folio, the housekeeping board, and the food-and-beverage handoffs talk to each other without the night auditor copy-pasting. Marriott has not handed me an architecture diagram and I would not print it if they did, but the operator-visible change is consistent across the properties I have talked to: fewer tabs, faster lookups, a unified guest record that no longer requires the front desk to ask a returning Bonvoy member for information the company already has.

Bonvoy, by the way, is what makes this expensive to get wrong. Capuano put the number at “nearly 283 million members at the end of March.” That is the database the new ecosystem has to honor — every preference, every stay history, every status calculation, every co-branded credit card linkage — in real time, at the moment the front-desk associate is trying to upsell a club-level upgrade. Marriott runs 37 co-branded cards across 13 countries; the company disclosed that those co-branded credit card fees grew 37% in the quarter. That growth does not happen if the loyalty plumbing breaks.

The ecosystem also replaces — and this is the part the analysts on the call did not press on — a generation of bolt-on point solutions that owners had been paying for individually. Some of that owner spend goes away. Some of it gets reabsorbed into program fees. The net is something I will come back to in the section on cost absorption, because it is the part of the story franchisees are still working out at the kitchen table.

How owners are absorbing the cost

There is no clean public number for what a Marriott franchisee pays, per property, to run on the new stack. Owners do not publish it. The company does not break it out in a way that lets you back into it from the 10-Q. What I can tell you, from a half-dozen conversations across two brand families, is the rough shape.

First, there is a one-time conversion cost. Hardware refreshes where the existing terminals do not meet the new minimums. Cabling in some properties. Training time that comes off operating hours. The owners I spoke to described this as a “low five-figures to low six-figures” spend per property, depending on size and how much of the legacy stack was already current. None of them volunteered the number willingly. All of them, when pushed, said it was within tolerance for a flag of this size.

Second, there is the recurring fee structure. The new ecosystem is funded, in part, through the existing program fee — which is to say, owners are paying for it whether or not they would have chosen to. This is the part that produces the most heat in owner-advisory-council rooms, and the part Marriott will defend on the grounds that the program fee historically funds shared technology. The pushback, when I hear it, is not really about the principle. It is about the cadence: owners want the cost recognized on the same timeline as the demand-side benefit (more direct bookings, fewer OTA commissions, higher attach on ancillaries) actually shows up on the P&L. That is a timing argument, not an existential one.

Third — and this is the piece most operators underweight — there is the cost they are no longer paying. The bolt-on revenue-management module that the new system subsumes. The third-party guest-messaging tool that gets retired. The reporting subscription that nobody renews after the dashboard inside the new ecosystem hits parity. None of that shows up in a press release. All of it shows up in the property-level budget six to nine months after cutover.

The owners I trust most on this — the ones who have been through a flag conversion before, who do not get excited about brand standards announcements either way — describe the net as “neutral to slightly positive in year one, materially positive in year two if direct booking attach holds.” That is the bet. Marriott is asking owners to underwrite a stack that pays back through demand-side share gains, and the company is asking analysts to underwrite the same story through residential branding fees, co-branded credit card fees, and the gross fee line.

Natural-language search, on schedule

The part of the May 6 call that the trade press picked up — Hotel Dive ran it that afternoon — was Capuano’s commitment that natural-language search on Marriott.com and the app rolls out by the end of Q2 2026. Phased, not big-bang. End of June, in other words, give or take a week.

This is the consumer-visible artifact of the new ecosystem. It is the thing a Bonvoy member will type into a search box that does not look like the search box she remembers — “find me a beach hotel under three hours from Atlanta with a kids’ club and adjoining rooms for next President’s Day weekend” — and get a result that is not a brittle filter cascade. The plumbing for that capability is the same plumbing the front desk is now using. The unified guest record, the rate inventory surface, the loyalty engine, the property attribute taxonomy: those are not separate projects. They are the new ecosystem, viewed from a different angle.

The reason this matters for an operator case study is that the search rollout is the first public, gradeable proof point. The 1,000-hotel cutover is internal. Capuano can claim it; analysts can take or leave it on faith. Natural-language search lands in front of 283 million Bonvoy members and either works or does not. Conversion rate on direct-booking traffic in Q3 is the number that tells you whether the back-end investment is throwing off front-end yield. If you are an owner, you will know by the September STR report. If you are an analyst, you will know on the August Q2 call.

Marriott has been disciplined about not over-promising on the AI surface — Capuano did not pitch this as ChatGPT-for-hotels on the call, and the company has not put a flashy product name on it. That restraint is, I think, deliberate. The Hilton AI Planner has a name and a press release. Marriott’s natural-language search is launching as a feature inside the existing surfaces. Which brings me to the comparison.

Where Hilton went different — and why

Two weeks ago I sat with the Hilton side of this on the April 28 call. Chris Nassetta, asked about the AI Planner, named his vendor on the record: “we did with Anthropic and Claude.” He also went out of his way to remind the analyst that Hilton has 41 AI use cases in test, 9,100+ properties, more than 1.3 million rooms, and roughly 90% of its enterprise technology on cloud — up from about 20% in 2020. The headline soundbite, though, was the one about distribution power: “we are the only ones with that 25% of the market that can control rate inventory availability.” That is a different kind of moat statement. It is not about who builds the better assistant. It is about who controls the inventory the assistant queries.

Hilton’s posture is vendor-partner-stack. Anthropic for the consumer-facing AI Planner. Google and OpenAI elsewhere in the 41-use-case portfolio. Hilton is not building the foundation models. Hilton is composing on top of them, fast, and using its cloud migration as the substrate that lets that composition happen at enterprise scale.

Marriott’s posture is build-with-vendors. The new ecosystem is proprietary in the sense that Marriott owns the integration layer, the data layer, the workflow surfaces, and the consumer experience. It rents the underlying cloud and the underlying LLMs. It does not, as far as I can tell from public commentary, hand a vendor the keys to the consumer surface the way Hilton handed Anthropic the AI Planner badge.

Both can be right. Hilton’s bet is that speed-to-deployment matters more than ownership of the surface — that 41 use cases in test today beats 4 use cases in production next year. Marriott’s bet is that owning the ecosystem produces stickier monetization downstream — residential branding fees, co-branded card fees, ancillary attach — because the ecosystem is the thing the brand is selling to the next residential developer or card-partner negotiator.

The fee mix tells you which bet is currently producing, and that is the next section.

What the fee mix tells you

Marriott’s Q1 gross fees were $1.43 billion, up 12% year over year. Inside that line, two components ran much hotter than the headline:

• Residential branding fees: +70%. This is the line that says luxury and ultra-luxury condo developers are paying Marriott more, faster, to put a Ritz-Carlton or St. Regis or W flag on a tower. Those deals do not close in a quarter. They close because a developer ran the math two and three years ago and decided the Marriott brand-and-platform combination was worth the fee. The platform piece of that math is the ecosystem.
• Co-branded credit card fees: +37%. Thirty-seven cards across thirteen countries. The card economics depend on Bonvoy engagement, which depends on the loyalty engine, which depends on the stack the company just cut 1,000 hotels over to. A 37% lift on a line of that size is not seasonality. It is platform-driven.

The company’s net rooms growth is 4.5%, with a 618,000-room pipeline behind it. Pipeline of that size means the fee line has a multi-year runway even before you factor in the platform-driven uplift. Adjusted EBITDA at $1.4B, up 15%, on a single-digit RevPAR comp tells you the operating leverage is real.

Hilton’s fee mix will look different at the line level when it reports the same way, because Hilton’s monetization story is more weighted to net unit growth and management/franchise fees than to the residential-and-card combo that Marriott is over-indexing on. Neither is wrong. They are just expressions of different platform choices.

If you are reading this as an operator — independent, franchisee, third-party manager — the fee mix is the thing to anchor on. Marriott’s platform is monetizing in lines that are not the room-night fee line. Which means the case for the platform does not depend on this quarter’s RevPAR comp. It depends on whether residential developers and card partners keep underwriting it. They did this quarter. The question is whether the natural-language search rollout, and the operational uplift the new ecosystem produces at property level, keeps them underwriting it for the next eight quarters.

What an independent or franchisee should copy

I cover this column for the operator who does not run 9,400 hotels. The right question is not “should I build a 1,000-hotel ecosystem.” The right question is “what part of what Marriott just did is something I can run a smaller version of, this quarter, in my property or my portfolio.”

Three things travel.

First: consolidate the surfaces your front desk actually touches. You will not build a Marriott-grade integration layer. You can, this quarter, audit which tabs your night auditor has open at 6 a.m. and ask which two could be retired or merged through configuration of tools you already pay for. The cost is staff hours. The payoff is recovered minutes per shift, which roll up into a real number across a year. The voice-agent maturity-curve piece we ran earlier this year walked through how to do that audit without getting talked into a stack you do not need yet.

Second: make the guest record portable inside your four walls before you worry about anything outside them. Marriott’s bet is that a unified guest record produces ancillary attach. The same bet works at property scale. If your F&B system does not know who your most-engaged repeat guest is when she sits down for breakfast, that is the integration to fix before you fix anything else. Our earlier piece on OpenTable’s AI surface covered the F&B-side version of this with examples that translate.

Third: decide whether you are building or composing. Marriott is building. Hilton is composing. The right answer for a 14-room boutique or a 22-property regional franchisee is almost always composing — pick the surfaces, pick the vendors, integrate the data, do not try to own the layer underneath. The Marriott AI deployment piece we ran in March walked through what “composing intelligently” looks like in practice for an operator who does not have a CTO.

The unifying point: the strategic frame is not who has the better AI tool. It is whether your tech choices produce fee economics — direct booking attach, ancillary spend, repeat visit rate, loyalty-driven ADR premium — that you can point to in a P&L six months from now.

Operator takeaways

• The 1,000-hotel number is the proof the playbook scales property-by-property; the next 8,000 properties are pacing, not engineering risk.
• Owner cost absorption is real in year one, neutral-to-positive in year two if direct booking attach holds. Underwrite the year-two number, not the year-one number.
• The fee mix — residential branding +70%, co-branded credit cards +37% — is what tells you the platform is monetizing. RevPAR is not the right scoreboard for this story.
• Natural-language search ships by end of Q2. Conversion lift on direct-booking traffic in Q3 is the first public, gradeable proof point.
• If you are not Marriott or Hilton, compose; do not build. Consolidate surfaces, unify the guest record, then layer AI on top of clean data.

The Q2 grade

I will be back on this in August. The grade on the May 6 quarter is not the May 6 quarter. It is the next one. Three things to watch when Marriott reports Q2:

One, did the natural-language search rollout actually ship on time, and did the company put a usage number against it. Capuano committed to end-of-Q2. Phased rollouts can mean anything from “live for 5% of traffic on a single brand” to “fully live across Marriott.com and the app for all Bonvoy members.” The disclosure on the August call will tell you which.

Two, did the residential branding fee line hold its growth rate. A 70% lift is not annualizable forever, but a sequential softening would tell you the platform-driven developer interest is more cyclical than structural. A second hot quarter would tell you the opposite.

Three, did the cutover pace accelerate. A thousand hotels by May 5 is a milestone. Two thousand by mid-August would tell you the operational machine for converting properties is running ahead of plan. Anything less than 1,400 would prompt the question of whether the back-half plan slipped.

Hilton’s Q2 call will land in the same window. Nassetta’s update on the 41 use cases — how many moved from test to production, what the AI Planner conversion data looks like, whether the rate-inventory-control moat held against booking-tool agents — is the comparison set. That is the call where the strategic frame either holds or breaks.

The night auditor I started this piece with does not care about any of this. She cares about whether the system is faster than the one she learned on. Eleven days in, her honest answer is: “Ask me in a month.” That is, I think, the right answer for everyone reading this too.

— Priya files The Operator. Tips: tips@tabletransfers.com.