Will AI replace waiters? What the job data and diner surveys say

Not on current evidence. US waiter jobs are projected to grow 2% from 2025 to 2035, with about 423,100 openings every year (BLS, 2026), and most sit-down diners say they still want people serving them. What AI is changing is parts of the job: taking the order, answering the same menu questions again and again, and passing on calls and check requests. Carrying food, making the call on an allergy, reading a table and fixing a bad night stay with staff.

We make WaiterPilot, an AI waiter by BroadNet.AI, so we have a stake in this question. Our product is built to take order-taking off servers, not to replace them. That’s why the figures below come from government statistics, trade surveys and academic research, each with its source and date, and why we label vendor figures as such.

What does the job data say about waiters?

Short answer: the job is large, still projected to grow slowly, and hard to keep staffed.

US waiter and server jobs, the latest figures as published by October 2026:

Two things stand out. Growth is slow but positive: the BLS projection is not a forecast of decline. And the openings figure is huge compared with the growth, because most openings replace people who leave. Restaurants don’t have a “too many waiters” problem; they have a “can’t keep waiters” problem. A projection is not a promise, and it will be revised as tools change, but there is no sign in the official data of the job disappearing.

For more on staffing, see our restaurant labor shortage data for 2026.

Do diners want a person or a machine?

Short answer: most sit-down diners still want people, but younger guests are increasingly happy to order from a screen or an AI.

Read together, these say the same thing: guests accept technology for parts of the meal, as long as a person is within reach. A restaurant that removed its servers would be going against what most of its guests say they want.

Which waiter tasks can AI take, and which can’t it?

Short answer: AI can take the order-taking and the repeat questions; the physical work and the judgment stay with people.

This is the honest split, task by task, with what WaiterPilot does where it applies:

The pattern is clear. Software is good at the parts that repeat: writing down the same order, answering the same question about what comes with the burger, walking to the till. People are good at the parts that need judgment and hands. For a practical list of what to move off a server’s plate, see nine jobs you can take off your servers.

What about robot waiters?

Short answer: they are a different product, and we have no independent evidence that they replace servers.

Robot waiters are machines that carry plates from the kitchen to the table. They don’t take orders by conversation, check on guests or handle problems. We haven’t evaluated any, so we make no claims about them. The three kinds of product sold as an “AI waiter” are explained in AI waiter vs robot waiter vs voice AI.

Why do so many people expect AI to replace waiters?

Short answer: because the most visible restaurant AI, at drive-thrus and on phones, does take the whole order, and that gets mixed up with table service.

At a drive-thru, the order is the whole interaction: a voice at a speaker, a payment at the window, food in a bag. When voice AI takes that order, a job really does change. Table service is different. The order is a few minutes of a meal that can last an hour, and most of what a server does happens before and after it. Headlines about robots carrying plates add to the confusion.

Actual use is also smaller than the coverage suggests. In the National Restaurant Association’s 2026 report, only 6% of US restaurants used AI for customer orders, even though about a quarter used some AI-related tool (NRA, reported by Restaurant Dive, 2026). For a map of where AI is used today, see how AI is used in restaurants.

Will restaurants need fewer servers?

Short answer: the evidence shows faster meals and less time per table, not fewer servers, and the biggest time figure comes from one restaurant.

The strongest study is a peer-reviewed analysis of a 66-restaurant US casual chain. Tabletop ordering devices were linked to meals about 10% shorter and about 1% more per check (Tan and Netessine, Management Science, reported by Restaurant Technology News, 2019). Shorter meals mean more tables served in an evening by the same team, not a smaller team.

A Cornell report found server time per table fell 36% when guests paid by device, and 65% when they ordered and paid by device, from 10 minutes 16 seconds to 3 minutes 30 seconds (Cornell Hospitality Report, 2016). That was one restaurant, with 7 tables in the order-and-pay group, and a co-author worked for the device vendor. It is a hint, not a rule.

Operators themselves don’t expect replacement: 69% said technology “will augment rather than replace human labor” (NRA, 2024). We have no data showing WaiterPilot reduces headcount, and we don’t sell it that way.

What does a server’s evening look like with an AI waiter?

Short answer: fewer trips to take orders and fetch checks, and more time at tables that need a person.

Casa Olivo is a fictional restaurant (an example, not a customer). The minutes below are our assumptions, not measurements. Time your own service before relying on any of them.

Assumptions for one server’s section on a Friday: 6 tables, each ordering twice (mains, then dessert or drinks), 3 minutes per order-taking visit, and 2 minutes per check run.

On these assumptions, the server gets back about half an hour in a section, while carrying every plate, answering every waiter call and handling every allergy question in person. Some guests will still want to order from the server, and they should be able to. The table also shows what doesn’t move: the check still needs a person, because WaiterPilot takes no payment.

Are waiters and servers the same?

Short answer: yes. “Server” is the gender-neutral name for the same job.

The US Bureau of Labor Statistics still calls the occupation “waiters and waitresses”. Most restaurants now write “server” on schedules and job ads. In this article we use both to mean the same job.

What should owners and servers do now?

Short answer: owners should move repetitive tasks to tools and keep people on hospitality; servers should lean into the parts software can’t do.

For owners:

1. List the tasks your servers repeat most on a busy night.
2. Move one of them to a tool, and keep a human path for guests who want it.
3. Keep allergy decisions, complaints and payment with staff.
4. Measure time to first order and time to check before and after. Don’t expect bigger checks: the peer-reviewed effect is about 1%.

For servers: the valuable skills are the ones in the bottom half of the task table. Knowing the menu deeply, handling allergies with care, reading a table and turning a bad night around are what guests come back for.

What to do next

If your servers spend their evenings writing down orders and walking to the till, see how an AI waiter takes that part over while staff keep the rest. Try a sample order from the guest side, and check pricing: $200 per location per month after 30 days free, with no contract.