Service Robots
Delivery robots are sold as one category and used as three jobs. One carries sealed items down a hotel corridor. One runs plates in a dining room. One is really a screen that happens to roll. If you mix those jobs, you will measure the wrong building and buy the wrong machine.
This is a selection guide, not a payback spreadsheet. For hotel labour maths see whether hotel robots pay off. For dining rooms see delivery robot versus human service. Here the question is which type you should even shortlist.
ⓘ Quick Answer
Choose an amenity delivery robot such as Relay for indoor, repeatable, closed-compartment runs in hotels, campuses, and similar buildings with lift integration if you are multi-floor. Choose a food runner only if aisles, payload, and a hybrid with staff make sense. Choose Temi if the payload is a video call, not a bag. Choose Pepper if the job is greeting, not carrying. Outdoor last-mile sidewalk robots are a different procurement and are not what these indoor machines do.
Name the payload and the route
Write one sentence: what moves, from where, to where, how often, and who loads it. Amenities to guest rooms, plates from pass to table, documents between floors, and “the robot should also say hello” are four sentences. They are not one robot.
- Closed compartment, staff at both ends: classic indoor delivery. Relay-class machines live here.
- Open trays of food: restaurant running. Layout and spill risk dominate. Headcount usually stays.
- A screen that must arrive: telepresence. That is Temi, not a delivery bay.
- A face at the door: Pepper. It does not replace the delivery fleet.
Indoor versus everything else
The machines on Robot.help in this cluster are indoor, flat-floor devices. Sidewalk last-mile, road crossing, and weather are another industry. If your route includes public pavement, do not force a hotel robot into it. If your route is corridors and a lift, do not buy a sidewalk unit “for future flexibility”.
Multi-floor indoor delivery almost always means electronic lift control. That integration is often the largest hidden cost and the most common reason projects stall. Walk the route with a loaded trolley. Note the narrowest gap at peak, every threshold, and every door that does not stay open.
💡 What to Check
Wi-Fi along the whole path, including lift shafts if relevant. A dock that does not block egress. A stuck-robot procedure owned by a named role, not “the night team will figure it out”.
Payload, volume, and concurrency
Weigh the heaviest real item, not the brochure item. Include packaging. Measure the compartment against your tallest bag and your fullest tray. Then count concurrent trips in the busiest thirty minutes. One robot makes one trip at a time. Check-in rushes and dinner passes are where fleets are sized, not daily averages.
Matching catalog machines without pretending they are clones
Relay
Indoor amenity and supply delivery. Judge mapping, compartment, lift support, and vendor deployment — not cuteness.
Temi
Rolling screen, voice, follow. Judge it as telepresence. It is a weak choice if the job is moving towels.
Pepper
Social greeter with a screen. Use it at the door if greeting is the job. Do not ask it to be a dumbwaiter.
Hospital logistics such as Moxi are closer to delivery than to greeting, but they are clinical programmes with infection-control rules. Do not treat a hotel robot and a hospital assistant as interchangeable because both have wheels.
Operations beat the unboxing
Who loads, who receives, who cleans trays, who answers when it stops in a doorway, and what happens during charging. Software subscription and local support belong in the quote. A pilot on one floor with one payload is how you learn. A building-wide launch on day one is how you buy a story for the newsletter and a problem for the duty manager.
- Define payload, route, and peak concurrency.
- Prove the floor and lifts with a trolley, then with a vendor demo on your site.
- Quote the integration, not only the robot.
- Run a time-boxed pilot with a named owner.
- Expand only on routes that recovered walking time or service quality.
Selection checklist
Do not sign until you can answer
- Is this indoor amenity delivery, food running, telepresence, or greeting?
- What is the narrowest point on the busiest route, measured at peak?
- Do we need lift integration, and who pays for it?
- What is the real payload, including packaging, and how many concurrent trips?
- Who loads, unloads, and rescues the robot on a Sunday night?
- Is local support available in our country?
Final Thoughts
Choose a delivery robot the way you choose a trolley: by what it carries and where it fits. Compare Relay, Temi, and Pepper on Robot.help as three answers, then cost the building, not the keynote.
Frequently Asked Questions
Can one robot do hotel delivery and restaurant running?
Sometimes vendors will say yes. Layouts rarely agree. Test the harder route. Food and sealed amenities are different spill and hygiene problems.
Do we need outdoor capability “just in case”?
Not if your actual work is indoors. Outdoor autonomy is a different safety and regulatory problem.
How many robots do we need?
Size on peak concurrent trips, not daily totals. One machine serves a trickle and queues a rush.
Is Pepper a delivery robot?
No. It is a social service robot. If the payload is physical, start with a delivery platform.
Should we start with ROI or with layout?
Layout first. A robot that cannot complete the route has an infinite payback period. Then use a walking-time framework if the route works.