AI Receptionist for Restaurants: Reservations, Orders and Limits

At 7:15 on a Friday, the phone at most independent restaurants rings into an empty host stand. The host is seating a four-top, the manager is expediting, and whoever is nearest the phone is carrying three plates. The call goes to voicemail, and nobody listens to voicemail until Monday.

Some of those calls were "are you open on Sunday?" One was a party of twelve for next Saturday. One was an office that wanted catering for forty. That is the problem an AI receptionist solves for a restaurant, and it is narrower than the pitch most vendors give.

Why restaurants are different

If you have read our pieces on clinics, dental offices, law firms and HVAC companies, the cost structure is the same. Three things change for a restaurant.

Most calls are small, and a few are worth a month. Hours, parking, "do you have vegan options?" add up to most of your call log and are worth almost nothing individually. Mixed in are the catering order and the large party, which can be worth more than a week of takeout. The job is to answer the first kind instantly and never lose the second kind.

The calls arrive exactly when nobody can answer. A clinic misses calls at lunch. A restaurant misses them at the peak of service, when every person in the building is busy doing the thing that earns money.

There is already a better channel for orders. Unlike a dental office or a law firm, a restaurant usually has a way to take an order without a phone call: an ordering page or app, where the choices are buttons and the ticket reaches the kitchen complete. The AI's best move is often to send people there.

What it should and shouldn't handle

The allergy rule

This is the one line a restaurant AI must never cross.

It can read what your menu says: which dishes are marked as containing nuts, dairy or gluten. It must never tell a caller that a dish is safe for their allergy, because whether it is safe depends on your kitchen, your fryer and your prep line, and only a manager or a chef can speak to that. Write the line yourself, something like "for allergies, our manager will call you back before you order", and have the AI say it word for word.

Why it should rarely take the order

Vendors love to demo an AI taking a phone order, because it is impressive. For most restaurants it is the wrong first build.

Phone orders are where mistakes start: "two birria, one with no onion" arrives as two with onion. An AI that takes the order has the same problem in a new form. It has to read back every item and modifier, the caller is in a hurry, and a wrong ticket costs you a remake and a regular.

The cheaper move is to answer the question, then text the caller a link to your ordering page while they're still on the line. The choices are buttons, the ticket reaches the kitchen complete, and nobody has to transcribe anything. Our ordering app case study shows why required choices beat free-text notes.

Connecting to your systems

The same three levels apply as in every other industry:

1. Request only. The AI takes the reservation or catering details and sends them to your host or manager by text or email. Works with any system on day one.
2. Read availability. It can see open tables in your reservation system and offer real times, while a person confirms.
3. Book directly. It writes reservations into OpenTable, Resy, Toast Tables or similar. This only works if your system allows outside tools, and it is the version most likely to double-book a Saturday.

Start at level one. A host who opens a clean list of requests before service is already far ahead of a voicemail box.

What it costs

• The build: $1,200 to $4,000 for a narrow first version, set mostly by the integration level above.
• Usage: a few cents per conversation. Even a busy restaurant is usually a few dollars to a few tens of dollars a month.
• The recurring layer: the phone number and telephony provider, plus any fee your reservation system or POS charges for integration access.

The arithmetic, with a worked example

You need four numbers: missed calls per week, the share of those that would have become revenue, a conservative recovery rate (assume a third), and what that revenue is worth.

weekly recovered value =
    missed calls × revenue share × recovery rate × average value

Restaurants are lumpy, so run it twice: once for ordinary calls and once for catering and large parties. Both illustrations use made-up numbers, so substitute your own:

Ordinary calls. A restaurant misses 25 calls a week during service. About 30% would have become a reservation or a takeout order, at a hypothetical $60 each: 25 × 0.30 × ⅓ × $60 = $150 a week.

Catering and large parties. The same restaurant misses 2 of those inquiries a week, and half would have booked, at a hypothetical $600 each: 2 × 0.50 × ⅓ × $600 = $200 a week.

Together that is about $350 a week, so a $3,000 build pays for itself in roughly two months. Notice that two catering calls outweigh twenty-five ordinary ones. If you only ever get a catering inquiry once a month, the case gets much weaker.

When it is not worth it

• Your phone barely rings. Counter-service places with online ordering often get few calls worth catching.
• You're full every night. If you turn away walk-ins on weekends, more reservation requests make the waitlist longer, not the revenue bigger.
• Nobody reads the requests. A reservation request that sits unconfirmed is worse than voicemail, because the guest shows up expecting a table.

New to this, or still comparing? Our guide to AI receptionists for small business covers the ways to buy one, and answering service vs AI receptionist puts published prices side by side.

How to start

Cover the dinner rush and after-hours only. Let it answer the common questions, take reservation and catering requests, and text your ordering link, with the allergy line written by you. Read the transcripts every week for the first month. You will learn what your guests actually call about, which is rarely what you'd guess.

Running an auto repair shop instead? There the rush is at the counter at 7:30 AM: see AI receptionist for auto repair shops.

For the ordering side of the same problem, see our ordering app case study. For a working conversational agent and its handoff to a person, see the clinic AI assistant case study.

When you want it scoped to your restaurant, our practical AI service starts with one question: how many calls went to voicemail last Friday night?

Frequently asked questions

How much does an AI receptionist cost for a restaurant?

A narrow build that answers hours and menu questions, takes reservation requests, captures catering and large-party inquiries, and texts callers your online ordering link typically runs $1,200–$4,000 to set up, plus a few cents per conversation and your phone provider's monthly fee. Connecting it to your reservation system or POS adds cost depending on what access they allow.

Can an AI receptionist take phone orders for a restaurant?

It can, but for most restaurants it shouldn't be the first thing you build. Phone orders are where modifiers, substitutions and allergies get lost, and an AI reading a long order back is slower than a caller expects. Texting the caller a link to your online ordering page, where the choices are buttons, is cheaper and gets the order into the kitchen complete.

Can an AI receptionist book reservations on OpenTable, Resy or Toast?

It depends on what your reservation system allows outside tools to do, so check before anything is quoted. Many restaurants don't need it to start: an AI that takes the party size, date, time and phone number and sends it to your host to confirm captures most of the value with no integration at all.

Should an AI receptionist answer allergy questions?

Only with a fixed line you wrote, never with its own answer. It can say which dishes the menu marks as containing nuts or gluten, but it must never tell a caller a dish is safe for their allergy. Cross-contact happens in kitchens, and only a manager or chef can speak to that. The safest script hands those calls to a person or asks for a callback.

Is an AI receptionist worth it for a small restaurant?

Only if you actually miss calls that would have become revenue. A counter-service place with online ordering may get few calls worth catching. A restaurant that takes reservations, gets catering inquiries, or has one person answering the phone during the dinner rush usually has enough missed calls to make the numbers work.