Can robots work in hotels? Tasks, costs and real deployments

Every hotel manager who has seen a robot video eventually asks the same thing. Can a robot actually do useful work here, or is it another gadget that ends up in a storage room?

The honest answer sits between the two extremes. Robots in hotels work today for a few narrow jobs, mostly delivery and cleaning. They do not replace the front desk, and I would not try to make them.

If you want the wider picture of where this fits across other industries, I keep a hub on robots in business that covers the same ground outside hospitality.

The short answer

Yes, for a few narrow jobs. No, for replacing people at reception.

Japan's Henn na Hotel is the cautionary tale, and I will get to it below. The hotel went heavy on robots, then cut more than half of them. That story gets quoted a lot by people who want the whole category to fail.

The other side of the story is quieter. One delivery-robot maker, Relay Robotics, reports more than 1.5 million deliveries across hotels and hospitals, and it sells on reliability numbers, not novelty. That contrast is the real state of the market.

What robots actually do in hotels today

Three jobs account for nearly every real deployment.

Room delivery

A delivery robot carries amenities or a room service order to the guest room and notifies the guest when it arrives. This is what a hotel with robot room service actually looks like. That sounds simple. The hard part is the elevator.

For a robot to work above the ground floor, the lift has to talk to it. Relay Robotics states that during installation, elevators are configured to communicate with its robots, so the robot can call a lift, select a floor, and enter and exit without a person pressing buttons. Pudu's FlashBot Max is listed with KONE and OTIS elevator integration for the same reason.

If your building cannot support that integration, a multi-floor robot becomes a ground-floor robot. Ask about this before anything else.

Floor cleaning

Autonomous scrubbers clean lobbies, corridors and ballrooms overnight. This is the least glamorous job on the list and probably the easiest to justify, because the work happens when guests are asleep and nobody is watching.

Guiding and greeting

Guiding robots direct guests around the property. Guest-facing robots are where Henn na struggled most, and I think this is the weakest use case. Guests ask unpredictable questions. A robot that cannot answer them is a moving sign, and a sign costs a lot less.

What a hotel robot costs

Prices are list prices from one US reseller, RobotLAB, checked 15 October 2026.

Do the maths on any row and the monthly price is about 2.8% of the purchase price. Over 36 months you pay roughly what buying would cost. Renting is not cheaper. What it buys you is a smaller upfront bill and, if the contract says so, a vendor who has to keep the robot running.

When you compare quotes, ask what the monthly figure actually includes. Installation, elevator integration, maintenance and network setup are the four items I would want in writing. A cheap monthly rate that excludes elevator work is not cheap once the integration invoice arrives.

Relay Robotics does not publish pricing at all. It sells by custom quote, and the company says cost depends on deployment complexity, the number of floors and the number of robots. If you are pricing a multi-floor property, expect that conversation rather than a list price.

The hotel that fired its robots

In January 2019, Hotel Technology News reported that Japan's Henn na Hotel, owned by H.I.S., had cut over half of its 243-robot workforce. The details are brutal in a useful way.

The in-room assistant Churi could not answer basic guest questions and was among the first to go. Luggage-carrying robots could not handle their one job. A dinosaur robot could not scan guest passports. Dancing robot dogs often failed to dance.

Read that list again. Every failure was a robot doing something a guest could see and judge in real time. Nobody notices a scrubber working at 3am. Everybody notices a dinosaur that cannot read a passport.

My read: the lesson is not that robots fail in hotels. It is that guest-facing novelty fails, and back-of-house utility survives. Put the robot where the work is repetitive and the audience is small.

Where it is heading

The next test is bigger. YYForce plans to deploy service robots at Novotel Singapore on Stevens and Mercure Singapore on Stevens, a 772-room precinct, according to an announcement from 14 October 2026.

Be careful with the numbers here, because they are projections, not results. The confirmed first phase covers guest deliveries, food and beverage logistics, scheduled cleaning and concierge support. Lift integration, route mapping and network work were still under way at announcement. A second phase is intended to focus on YYForce's humanoid, informed by housekeeping and banquet workflows.

The company projects roughly 10% lower monthly operating costs in selected housekeeping, concierge and F&B operations, including robot costs. It also says actual costs will be evaluated after rollout, and it has not disclosed how many robots will be deployed. Treat the 10% as a target, not an outcome.

For scale on the delivery side, Relay Robotics reports over 1.5 million lifetime deliveries, more than 25,000 deliveries per month, and a 99.8% successful delivery rate, all company-reported. Its partner logos include Hyatt, Hilton, Sonesta, Marriott, Embassy Suites by Hilton and Delta Hotels. If you want one signal that a robot for hotel delivery is not experimental, that is it.

How to decide if your hotel needs one

Work through these before you take a demo.

1. How many deliveries happen per night? Count them for a week. If the number is low, a robot will sit idle and you will resent it.
2. Are your lifts compatible? Multi-floor delivery depends on elevator integration. If your lift vendor or building management will not cooperate, stop here.
3. Who handles the robot when it gets stuck? Someone has to retrieve it, reset it and deal with a guest who is waiting. Name that person before you buy.
4. Can you rent before you buy? A 36-month rental spreads the cost, and a good contract puts uptime on the vendor. I would rent first on any unproven workflow.
5. What does the vendor measure? Ask for successful delivery rate, interventions per week and uptime. If they cannot report those, you are buying hardware, not a service.
6. Does the job face guests or not? Back-of-house cleaning is a safer first project than a greeter in the lobby.

None of this is exotic. It is the same discipline you would apply to any equipment purchase, and it is the difference between a robot that earns its keep and one that ends up beside the luggage cart.

I cover which robots are actually earning their keep every Tuesday in the Physical AI Briefing. Get it by email.