Why table service keeps a person in it
Ask whether AI will replace waitresses and you are really asking about two different jobs inside one role. One part is clerical: ringing in orders, splitting checks, taking payment. That part has been moving to screens for years. The other part is physical and social: carrying hot plates through a crowded room, noticing the table that has gone quiet, fixing a wrong order before the guest asks twice. Software has made almost no dent in that second part.
Restaurants also run on judgment that nobody writes down. A server decides when to interrupt a conversation, which child needs a napkin, how to pace a four-course meal so the kitchen is not slammed. When a dish comes out wrong, the recovery is a conversation, not a refund button. Those calls happen in seconds, with partial information, in front of a paying guest.
Scale is part of the story too. This is one of the largest occupations in the country, with about 2,270,910 jobs and median pay of $35,230 a year (BLS, 2025). Low wages weaken the business case for expensive hardware, which is why most restaurant sector automation so far has been cheap and fixed: a kiosk by the door, a tablet on the table, a handheld terminal in an apron.
What software does, what it assists with, what stays with servers
The clerical slice is the most automated. Entering an order into the point-of-sale, firing it to the kitchen display, calculating and splitting a check, taking card or phone payment, and recording who ordered what are all handled by systems that already exist, often operated by the guest instead of the server. The share of task time we score as work AI can handle on its own is 0%. The coverage method page explains how that share is built from task time.
A second group is assistance rather than handover. Describing specials and ingredients, answering allergen questions, and suggesting a drink with a dish can all be prompted by a tablet or an earpiece, but a person still delivers the answer and reads whether the guest wants more detail. Reservation and waitlist systems work the same way: they manage the queue, the server manages the room. Task time where AI supports the work instead of taking it is 27%.
What is left is the bulk of a shift. Serving food and beverages at the table, checking back during the meal, clearing and resetting covers, handling a complaint, and reading a table’s mood sit with people. Tray-carrying runners on fixed routes can move plates between the kitchen and a station, but someone still places the dish, answers the question that follows, and deals with the spill. The share of task time that still needs a person is 73%.
What has actually been tested
Not much, and that matters. Our evidence grade for how well AI performs against a qualified server is D, which means there is no published head-to-head test of AI against people doing this job. The benchmarks that exist measure text and office tasks, not whether a machine can carry four entrees across a dining room and keep a table happy.
So we give no parity number here. We will not invent one. What would settle it is a dated, published comparison inside working restaurants: machine-served versus staff-served sections measured on order accuracy, timing, guest satisfaction and labor hours, over a full service period rather than a launch week. Until something like that exists, claims that most serving work could be handled by robots are forecasts, not results. The method we use to score jobs sets out how grades change when real tests appear.
When the picture could shift
Most likely after 2044 (8 in 10 of our scenarios). The replacement-year method explains what that window is measuring.
Two things could pull it earlier. Cheaper mobile robots that handle uneven floors, chairs and children would turn food running from a novelty into standard equipment. And counter-service formats keep growing: if more meals are ordered on a screen and picked up at a window, fewer tables need a server at all, whatever the technology can do. Our guide to robots and physical jobs covers how slowly that hardware has moved.
Two things hold it back. The automation on offer today is mostly fixed in place, so it cannot cover the messy middle of a shift, and dining rooms are built for people, not machines. And guests push back: hospitality is part of what they pay for, and several operators have walked back heavy kiosk and chatbot setups. The federal outlook reflects a steady rather than shrinking role, with about 2% employment growth projected for waiters and waitresses from 2025 to 2035 (BLS, 2025).
How servers stay needed
Lean into the parts of the shift a screen cannot do. Table service itself, from pacing courses to reading when a guest wants to be left alone. Recovery, where a wrong or late dish becomes a guest who comes back. And menu depth, the kind of answer that turns a question about a dish into a sale the kitchen is happy to make.
Two skills travel well from here. Beverage knowledge, since wine, cocktail and coffee programs pay more and lean on taste and conversation. And floor leadership, meaning the ability to run a section, train new hires and keep service moving when the system goes down.
What to do: learn the POS and ordering tech well enough to fix it mid-service, because the server who keeps the room running when the kiosk freezes is the one the manager keeps.
Nearby jobs score differently, and the reasons are worth reading side by side. The closest are bartenders, food servers, nonrestaurant and fast food and counter workers. You can also see the wider food and beverage serving workers family, put two roles next to each other on the compare page, or browse the list of jobs that mostly need a person if you are weighing a move.