Not Every Guest Question Deserves a Human: Hotels Are Redrawing the Line Between AI and Staff

A guest calling down at two in the morning to ask for extra pillows does not want a warm conversation. They want pillows. That small, almost embarrassing observation is becoming one of the more useful ways to think about artificial intelligence in hospitality: not as a threat to service, but as a filter that separates the requests that need a human from the ones that never really did.

For years the industry treated automation and hospitality as opposites, as if every self-check-in kiosk chipped away at the soul of the business. That framing is losing ground fast. The conversation happening inside hotel groups right now is not whether to deploy AI at the front desk, but which slice of the guest journey should keep a human voice attached to it, and which slice was only ever administrative work dressed up as service.

The check-in test

At the Smart Travel News Roadshow in Barcelona this March, Idoia Herrero, HiJiffy's senior country lead for Spain, used the traditional check-in to make the point concrete. Scanning a document, waiting for a system to load, repeating the same basic instructions, filling out a form: none of that requires empathy or judgment. It requires speed and accuracy, and in the meantime it delays a tired traveler from the one thing they actually came for, which is their room. Herrero's framing was blunt: automate the transactional, and use the time it frees up to humanize what is left.

That reframing matters because hotels are not really choosing between AI and a fully staffed front desk. They are choosing between AI and an understaffed one. Hospitality loses an estimated 70% to 80% of its US workforce every year, according to turnover research cited by hotel technology firm TrustYou, and the UK's hospitality sector posts the highest average staff turnover of any industry, around 52%, per the Office for National Statistics. Recruiting alone will not close that gap. It has become a background condition that every technology decision in the sector now has to account for.

What Hilton actually automated

Hilton offers the clearest large-scale example of how this plays out in practice. The company now runs 41 distinct AI use cases across its 7,500 properties in 138 countries, according to CEO Chris Nassetta. Its guest-facing chatbots resolve routine questions, about hours, parking or amenities, roughly 50% faster than before and draw 90% positive guest feedback, freeing front-desk staff to handle more complex requests. The more telling investment, though, sits behind the scenes: Hilton built a generative AI coaching platform with SweetRush to train more than 400,000 employees in service recovery, the delicate skill of fixing a guest's bad moment before it becomes a bad review. Nassetta has described the philosophy behind it as enablement, not replacement. The company is not using AI to need fewer people. It is using AI to make the people it has better at the part of the job a machine cannot do.

Marriott took a more cautious route toward the same idea. Its recently deployed front-desk AI tool automates behind-the-scenes administrative work, like room assignment, rather than guest-facing conversation, a reminder that the line between what should be automated and what should stay human does not always fall in the same place twice.

Where the line actually falls

On the ground, the division looks less like a strategy slide and more like plain math. A large share of what hits a hotel's phone lines, room service orders, requests for extra towels, questions about breakfast hours, a taxi to the airport, is repetitive by nature and does not become more valuable for having a person read it out loud. Contler, the Bogotá-based voice AI company founded by Mateo Bolívar, has been testing exactly that boundary inside Colombian hotels: its agents processed more than 1,000 room service orders in a single week across a group of five-star properties, without a staff member on the line and without the errors or delays that come with an overworked night shift.

What is left for humans tends to be the moments carrying an emotional or reputational charge: the arrival of a guest coming off a 14-hour flight, an anniversary dinner that went wrong, a group booking with unusual requirements. When the Hilton Bogotá Corferias marked its seventh anniversary this year, it became the first hotel in Colombia certified to host Indian weddings, a high-value segment built almost entirely on personal trust and relationship management rather than transactional efficiency. AI did not win that certification. People did, with more time freed up to pursue it.

A staffing strategy, not a headcount cut

The practical risk for hoteliers is treating this shift as a cost-cutting exercise rather than a role redesign. Profitroom, the hospitality booking platform, put it plainly in a recent industry note: reception and reservation teams spend a disproportionate amount of their day answering the same handful of questions, and offloading that volume does not remove warmth from a property, it creates room for it. The job description for a front-desk employee is quietly shifting from administrative processing toward relationship management, upselling and service recovery, the tasks that actually shape whether a guest returns or leaves a review that hurts future bookings.

None of this settles the debate over exactly where AI's limits should sit; even Herrero admitted the industry has no fixed answer yet, only a shared intuition. But the direction is clear enough that hotels can no longer treat the question as hypothetical. The more useful exercise, at this point, is not asking whether AI belongs in a hotel. It is walking through a single day of guest requests and asking, honestly, which ones actually needed one of your people, and which ones were simply taking their time away from the guests who did.