Restaurant AI receptionist: answer the phone without leaving the floor

It's 19:40 on a Friday. Every table is turned or about to be, there are four walk-ins at the door, the pass is calling for runners and the phone by the host stand has rung three times in five minutes. One is a table for two tomorrow. One is someone asking whether you do a gluten-free pizza base. One is a supplier. You can't tell which is which until you pick up, and if you pick up, the walk-ins wait.

A restaurant AI receptionist answers that phone so the person on the door doesn't have to. It takes the booking, answers the question about the menu from the information you gave it, and leaves anything that needs a human in an inbox your team checks when the rush is over. This page explains how that works with Klugon, where it falls short, and how to set it up so it helps on a busy service rather than adding another thing to manage.

We've kept this practical. There are two sample conversations, a setup checklist, the allergen rules you have to get right, and costs worked through with the assumptions written down so you can swap in your own numbers.

What a restaurant AI receptionist actually does

Strip away the marketing and the job is simple. The AI answers when you can't, says the same correct thing every time, books what it's allowed to book, and writes down what it isn't.

With Klugon, that covers four places guests reach you:

• The phone. Your number forwards to a Klugon number, either for every call or only for the ones you don't pick up. The AI talks to the caller in a natural, AI-generated voice and every call appears in your inbox afterwards. The AI phone receptionist page goes into how calls are handled in detail.
• WhatsApp. Many guests would rather message than ring, particularly for anything that isn't urgent. The WhatsApp AI chatbot answers on your own WhatsApp Business number.
• Instagram and Facebook Messenger. Restaurants get a surprising number of booking requests as replies to stories. The same assistant can answer those, after the Meta setup described further down.
• Your website. A chat window you add with one line of code, plus a public page with your menu and a QR code you can print for the window or the bill folder.

All four feed one inbox, so the person who opens up on Saturday morning sees what came in overnight in one list rather than four apps.

It doesn't replace your host, your manager or your reservation book. It removes the moment where someone has to choose between the guest in front of them and the phone.

The calls and messages a restaurant gets

Before you set anything up, it helps to be honest about what actually comes in. Most restaurants find the same handful of call types make up nearly all of their volume, and only a few of them need a person.

Table bookings, changes and cancellations

The bulk of it. "Have you got a table for four on Saturday?" "Can we move our 7:30 to 8?" "We're running late, will you hold the table?" "We need to cancel, sorry." Booking calls are short, repetitive and time-sensitive. A guest who gets no answer at 19:45 on a Friday doesn't leave a voicemail; they book somewhere else.

The AI can check availability and book a slot. It can cancel a booking on the phone. Moving a booking to a different time works in chat and on WhatsApp, but on a phone call the AI can only cancel, so a caller who wants to move a table either rebooks or gets flagged for your team. That's worth knowing before you decide how much to send to the phone.

"We're running late" is a different kind of call. It needs someone who can see the room and decide whether to hold the table. The AI should take the name and the new arrival time as a note for the team, not make a promise.

Menu, allergen and dietary questions

"Do you have vegan options?" "Is the risotto gluten-free?" "My son has a sesame allergy, can he eat there?" "Do you do a kids' menu?" "Is the steak halal?"

These come in all day, and the honest answer depends on what's in your kitchen this week. The AI answers from the menu and allergen information you upload. If it isn't in there, it says so and flags the conversation. The allergen section below explains why this is the area where you want the AI to be most careful, and how to set it up so it never guesses.

Opening hours, location, parking and accessibility

"Are you open on Monday?" "Where do I park?" "Is there step-free access?" "Do you take card only?" "Can I bring my dog?" These are the calls that interrupt service most and need the least thought. They're the easiest win, as long as the information you give the AI is current. Holiday closures matter here: add closed dates in advance so the AI isn't telling people you're open on Boxing Day when you're not.

Takeaway and collection orders

If you do takeaway by phone, you'll know the pattern: orders cluster between about 17:30 and 20:00, and they arrive exactly when the kitchen is under most pressure. This is the call type where restaurants most need to think about what they want. An AI that reads out your menu is one thing. An AI that's supposed to put a paid order into your kitchen is another. The section on phone orders below sets out what Klugon does and doesn't do here, because it's different from the booking side.

Large groups, private hire and events

"We're looking for somewhere for 25 people for a birthday in March." "Do you do private hire on a Sunday?" "What's your Christmas party menu?" These are worth more than any single table and almost always need a conversation with your manager about minimum spend, set menus and dates. The AI's job is to collect the details properly and get them in front of the right person, not to quote a price it doesn't know.

Everything else

Lost property ("I left my scarf on Thursday"), complaints about last night, job applicants asking if you're hiring, suppliers chasing an order, the drinks rep, someone selling card machines, the local paper. None of these should be handled by an AI beyond taking a message, and you can tell it so.

When the calls come

The pattern is predictable in words even if the numbers differ for every site. Booking calls start in late morning and build through the afternoon. There's a lunchtime wave if you open for lunch. Then the heaviest stretch runs from about 18:00 until the second sitting is seated, which is exactly when nobody is free to answer. Late-night messages arrive after close, often from people planning the weekend from the sofa.

The year has its own shape:

• September to early December: Christmas party enquiries, many of them large groups that need a manager.
• Mid-February: Valentine's Day, when a single evening can be fully booked weeks ahead and every call is someone hoping for a cancellation.
• March: Mother's Day Sunday lunch, often the busiest lunch of the year.
• Spring and summer bank holidays, and the first warm weekends: calls about outdoor tables, dogs and whether you take walk-ins.
• January: quieter, with a lot of "are you open?" calls and deal-hunting.

Knowing this tells you when to update what the AI knows. If your Christmas menu isn't in its knowledge by the end of September, it will be telling party bookers it doesn't know.

What the AI handles well

Given good information, these are the jobs it's suited to:

• Booking a table into a free slot within the rules you set: how far ahead, how much notice, which times, which days you're closed.
• Cancelling a booking when a guest rings to cancel, which frees the slot for someone else.
• Moving a booking in chat and on WhatsApp.
• Answering menu, hours, location and policy questions from your own documents, in the same words every time.
• Collecting details for large groups and events so your manager gets a complete enquiry rather than a half-written note.
• Taking messages for anything outside its remit, with a name and number.
• Answering in other languages. Phone calls can be answered in English, German, French, Spanish, Italian, Portuguese, Dutch, Japanese and Arabic. That matters in tourist areas, where a visitor ringing from a hotel would rather book in their own language.
• Answering at 2am, when someone decides to book the anniversary dinner they forgot about.

It also gives you something most restaurants don't have: a record. Every call and message lands in the inbox, and the analytics show which hours are busiest and which questions the AI couldn't answer well. If it keeps getting asked about parking and doesn't know, that's a gap you can fill in two minutes.

What it must hand over to your team

Some calls should never be handled by an AI, however good it is. Decide these before you go live and write them into the assistant's instructions.

• Allergy questions it can't answer from your written allergen information. It should say it doesn't have that information and flag the conversation. It should never reassure a guest that a dish is safe.
• Complaints. A guest unhappy about last night wants a person. The AI should apologise, take details and flag it.
• Large parties and private hire above whatever size you set. It collects date, numbers, budget and contact details, then hands it over.
• Anything involving money: refunds, disputed bills, vouchers that didn't work.
• Late arrivals and requests to hold a table, because only someone in the room knows whether you can.
• Press, suppliers, job applicants and sales calls: take a message.
• Anything that sounds like an emergency. If someone reports feeling unwell after eating with you, that needs a manager the same day.

What "hand over" means in practice

This is the bit to understand clearly, because it's where people's expectations are often wrong. Klugon does not transfer a live call to a person. There's no "putting you through to the manager". How the hand-over reaches you depends on the channel:

• On WhatsApp, Instagram, Messenger and website chat, the AI marks the conversation as needing a person. It shows with a "Needs a person" banner in the inbox and, if you've connected them, is posted to a Slack channel or sent as a Zapier event you can route wherever you like.
• On a phone call, the reliable route is a message for the team: the AI saves the caller's name, number and a note of what they need as a lead. New leads appear on your leads list and post to Slack or Zapier in the same way, and the call itself is in your inbox.

It doesn't send email or push alerts to you. So the AI tells the caller something like: "I've passed this to the team and someone will call you back." Your manager sees it in Slack, the inbox or the leads list and calls back. On WhatsApp, Instagram and Messenger, your team can reply from the Klugon inbox, as long as it's within Meta's 24-hour reply window. On phone calls and website chat, they reply by calling or emailing the guest.

Two further details. First, the AI keeps replying after it flags a conversation; nothing pauses it. If a guest sends another WhatsApp message, the AI will answer it. Write your instructions so it doesn't promise things your team hasn't agreed. Second, someone has to own the inbox. If nobody checks the flagged conversations after service, the handover has gone nowhere. We cover handing conversations to staff in more depth elsewhere.

How table booking works in Klugon

This is where we want to be straight with you, because restaurant booking is different from booking a haircut and Klugon's booking engine was built around appointments.

Klugon books into its own built-in calendar. It uses your opening hours and time zone, a slot interval (every 15, 30 or 60 minutes, for example), how much notice a booking needs, how far ahead people can book, a buffer after each booking and your closed dates. Each bookable "service" has a length, and can be tied to named staff, with clashes checked per staff member. When a guest books, they get a booking reference in the same conversation.

What it doesn't have is a floor plan or a cover count. It doesn't know that table 6 seats four, or that you can push 3 and 4 together for a six. It thinks in slots.

Setting it up for tables

The way restaurants make this fit is to treat each table, or each group of identical tables, the way a salon treats a stylist:

1. Create a bookable service for each party size you take bookings for, such as "Table for 2, 90 minutes" and "Table for 4, 2 hours", with the length you turn tables in.
2. In each service, list the tables that can take it (T1, T2, T3...) where a salon would list stylists.
3. Set the slot interval to match your sittings, and a buffer if you need time to reset.
4. Add your closed dates and set a minimum notice so the AI doesn't book a table 20 minutes from now when the kitchen is already slammed.

Klugon then offers a time only if one of the listed tables is free for the full length, and allocates the first free one. It's not a full reservation system and you'll find edge cases, but for a small restaurant taking phone bookings it does the job. Keep large or unusual parties out of it entirely: tell the AI to collect the details and hand them over.

The AI appointment booking page explains the booking rules in more detail.

What happens after the booking

The guest gets their confirmation and reference in the conversation itself, on the call or in the chat. Klugon does not send a separate confirmation email, text or WhatsApp afterwards, and it doesn't send reminders. There are no deposits or card holds, and no no-show fees. If you rely on deposits to protect Saturday night, keep taking those through your existing system.

On a phone booking, the AI can't attach a note to the booking itself. If a caller mentions a birthday cake or a severe allergy, have the AI save it as a note for the team, which posts to Slack, so your team sees it before service. In chat and on WhatsApp, a note can be added to the booking.

New bookings can post to Slack or Zapier, so whoever runs the book sees them come in.

If you use OpenTable, ResDiary or another reservation system

Many restaurants already run their book through OpenTable, ResDiary, SevenRooms, Collins or similar. Klugon does not connect to any of them. It can't read their availability or write bookings into them.

That leaves three honest options:

• Klugon for enquiries only. The AI answers questions, takes booking requests as messages and flags them, and your team enters them into your reservation system. Fine if phone bookings are a small share and you mainly want the phone answered.
• Send people to your booking link. In chat and on WhatsApp, the AI can share a link from your links library, such as your online booking page. On the phone it can tell people where to book online. This keeps one source of truth.
• Hold a block of tables in Klugon. Some restaurants keep a set of tables for phone and walk-in trade anyway. You could put those tables into Klugon and leave the rest in your online system. It works, but you need discipline so the same table isn't sold twice.

Klugon can also check a Google, Outlook or Apple calendar for busy times and send a Calendly link, after setup and on a Business plan or higher, but that's not the same as two-way sync with a reservation platform. If deciding between tools is the real question, our OpenTable alternative comparison sets out where each one is the better fit.

Phone orders and takeaway

Klugon is not an ordering system. It has a product catalogue, so it can tell a caller what's on the menu, what's in each dish and what it costs, and it can suggest a side or a dessert. But on a phone call it can't take payment, and it doesn't send orders to your kitchen printer, till or EPOS.

If you take orders by phone today, you have two realistic set-ups:

• Point callers to your online ordering. The AI answers menu questions and then directs people to the ordering page you already use. In chat it can share the link directly.
• Take the order as a message. The AI writes down the order, name, number and collection time, saves it as a lead, which posts to Slack for the team, who check it and put it through the till. This needs someone watching Slack or the inbox during the takeaway rush. Some restaurants find that's a fair trade; others find it's worse than answering the phone.

If takeaway is most of your business, the takeaways page goes through this in much more detail.

Sample conversations

These are two examples of how conversations go once the assistant is set up. Both are samples written for this page. The restaurants and guests are made up, and the AI voice on a real call is AI-generated.

Sample 1: a booking call during Friday service

AI: Good evening, Osteria Lume. How can I help?

Caller: Hi, have you got a table for two tomorrow night, about eight?

AI: Let me check. For two tomorrow, Saturday, I can do 19:30 or 20:45. Eight o'clock itself is taken.

Caller: 19:30 is fine. Quick question, my partner has a nut allergy. Is that going to be a problem?

AI: Our allergen information shows that the pesto, the praline semifreddo and the almond tart contain tree nuts, and nothing on the current menu lists peanuts. I can't tell you a dish is safe for an allergy, though. Please tell your server when you arrive and they'll go through the menu with the kitchen. I'll also pass a note to the team so they know before service.

Caller: That's great, thanks.

AI: Can I take a name and a mobile number for the booking?

Caller: Priya, 07700 900456.

AI: Thanks, Priya. You're booked for two at 19:30 tomorrow, Saturday. Your reference is OL-4821. Is there anything else?

Caller: Is there parking nearby?

AI: There's a pay-and-display car park on Mill Lane, about two minutes' walk, and it's free after 18:00 on Saturdays.

Caller: Perfect. Thanks.

AI: See you tomorrow. Goodbye.

What happened behind the scenes: a two-person table booked into a free 90-minute slot, the booking listed in the dashboard and posted to Slack, and a note saved for the team, "Priya, nut allergy, booking OL-4821", which posted to Slack so the manager can brief the kitchen. The nut list came from the allergen matrix the restaurant uploaded. The AI was told never to describe a dish as safe, and it didn't.

Sample 2: a Christmas party enquiry on WhatsApp

Guest: Hi, do you do Christmas parties? We'd be about 18 people, early December

AI: Hi, yes we do. For groups over 12 we use our festive set menu, three courses. I've added the link to this year's menu below this message. Do you have a date in mind?

Guest: Friday 4th or Saturday 5th ideally. Is there a minimum spend?

AI: I don't have the minimum spend or private room details for those dates, so I'll pass this to Hannah, our events manager, who'll come back to you with options. Can I take a few details for her? Your name, the best contact number, roughly how many people, and whether anyone has dietary requirements?

Guest: Tom Reid, 07700 900812, 18 maybe 20, two vegetarians and one coeliac

AI: Thanks, Tom. I've passed this to Hannah with: Friday 4 or Saturday 5 December, 18 to 20 guests, two vegetarian, one coeliac. She's usually in from 10:00. The festive menu has a vegetarian option for each course, and the allergen sheet is on the same link.

Guest: Brilliant, thanks

What happened behind the scenes: a lead created with Tom's details and notes, the conversation flagged for the team, and a Slack post in the events channel. At 10:15 the next morning Hannah replies to Tom from the Klugon inbox on WhatsApp, within Meta's 24-hour window. The AI didn't quote a minimum spend or promise a date, because it hadn't been given either.

Allergen rules for restaurants (FSA)

This is the most important section on the page. If your AI receptionist gets a menu price wrong, you have an awkward conversation. If it gets an allergen wrong, someone can end up in hospital. The rules exist whether or not you use an AI; the question is how to make sure the AI stays inside them.

What the law requires

In England, Wales and Northern Ireland, the Food Standards Agency's allergen guidance for food businesses sets out the obligations. In short:

• There are 14 allergens you must declare when they're used as ingredients: celery, cereals containing gluten, crustaceans, eggs, fish, lupin, milk, molluscs, mustard, peanuts, sesame, soybeans, sulphur dioxide and sulphites (above 10 parts per million), and tree nuts.
• For food that isn't prepacked, such as a restaurant meal, you can give the information in writing, on the menu or a chart, or verbally, as long as there's a written notice telling customers how to get it.
• For distance selling, which includes food ordered by phone or online, allergen information has to be available at two points: before the order is completed and when the food is delivered or collected. Before the order, it can be written (website, menu) or spoken (by phone).
• Staff must be trained on allergens.

In February 2025 the FSA published best practice guidance on allergen information for non-prepacked foods. It recommends that written allergen information is always available, backed up by a conversation between the customer and staff. It's guidance rather than new law, but it tells you what the regulator considers good practice, and it fits well with an AI that answers from a written allergen matrix.

In Scotland, Food Standards Scotland is the regulator and publishes its own guidance. The 14 allergens are the same.

How to set up the AI so it never guesses

The AI answers questions by searching the documents you give it. That's useful here, because it means you control the source. Do this:

1. Upload your allergen matrix, the grid of dishes against the 14 allergens, as a PDF, Word document or CSV. Scanned images of a printed sheet won't be read, so use the original file.
2. Write the rule into the assistant's instructions in plain words. For example: "Only answer allergen questions from the allergen matrix. List which dishes contain the allergen. Never say a dish is safe, allergen-free or suitable for someone with an allergy. Always tell the guest to speak to their server, and flag the conversation."
3. Cover cross-contamination honestly. If you use nuts in the kitchen, put a line in your information saying so, and the AI will repeat it.
4. Update the matrix whenever the menu or a supplier changes. An uploaded file doesn't update itself. If you change the pesto recipe, re-upload. Make it part of the menu change routine, the same way you reprint the menu.
5. Check the "questions without a good answer" list on the insights page every week. Allergy questions the AI couldn't answer will show up there.
6. Test it yourself before going live, using the browser test, with the awkward questions: "Can my coeliac daughter eat the fish and chips?" "Is there sesame in the bread?" "Is it safe for a severe peanut allergy?"

One practical limit: when a guest mentions an allergy on a phone booking, the AI can't attach it as a note to the booking. Make the instruction "whenever a guest mentions an allergy on a call, save their name, number and the allergy as a note for the team". The note posts to Slack, so it reaches your team before the guest arrives.

The AI doesn't replace your written allergen information on the premises or the conversation your staff have at the table. It's one more place where the same written information is repeated accurately.

Customer data and AI calls

Every booking involves a name and a phone number, and sometimes health information such as an allergy. In the UK, the Information Commissioner's Office guidance for organisations covers what you need to tell customers about how you use their data.

Klugon is operated by a German company under the GDPR. Its sub-processors are listed in the privacy policy, with transfers covered by standard contractual clauses. Calls are not recorded; every call appears in your inbox as a conversation. You can export your data as CSV and request full deletion by email. Mention Klugon in your own privacy notice as a service you use to answer calls and messages.

The default greeting doesn't say the caller is talking to an AI. Many restaurants prefer to write their own greeting that does, such as "Hi, you're through to Lume's AI assistant, how can I help?" It's a choice worth making deliberately.

Setup checklist for a restaurant

Here's the order we'd do it in. Most of the effort goes into the information, not the software.

Before you start

• Your current menu, as a PDF or text, with prices.
• Your allergen matrix, as a PDF, Word file or CSV (not a photo).
• Opening hours for each day, and the dates you're closed over the next few months.
• Your booking rules: party sizes you'll take, how long you turn tables, latest booking time, minimum notice, how far ahead.
• Your policies in plain words: dogs, children, corkage, cake fees, card only, service charge, dress code.
• A list of what the AI must always hand over: complaints, groups over a set size, private hire, allergy questions it can't answer, anything about money.
• Who owns the inbox during and after service.

In Klugon

1. Sign up and start the 7-day free trial on the voice product (a card is needed, and it's charged when the trial ends unless you cancel).
2. Fill in your business profile and opening hours. Set your time zone.
3. Upload the menu and allergen matrix to the knowledge base. Paste your policies. Or run the website scan, which reads your homepage and up to seven key pages.
4. Add your links: online booking page, takeaway ordering page, festive menu, map.
5. Create your table services and list the tables in each, as described above. Set the slot interval, buffer, minimum notice, booking window and closed dates.
6. Write the assistant's instructions: tone, greeting, allergen rule, hand-over rules.
7. Pick a voice and a language.
8. Test it in the browser. Ring it with the questions guests actually ask. Try to break it.
9. Connect Slack or Zapier so flagged conversations and new bookings reach whoever needs them.

Going live on each channel

• Phone. Phone numbers can only be added once the trial ends; during the trial you test in the browser. Then either buy a new number or keep your existing one: Klugon gives you a dedicated number and you set up call forwarding on your current line with your phone provider. Many restaurants forward only on no answer or busy, so staff still pick up when they can. We explain how to keep your number with call forwarding.
• WhatsApp. Connect a WhatsApp Business number through the Meta WhatsApp Cloud API. You'll need a Meta app, the phone number ID and an access token. Meta's WhatsApp Business site explains the account side.
• Instagram and Messenger. Connect your Facebook Page and professional Instagram account. Instagram messaging needs Meta app review before members of the public can message the assistant. Our page on Instagram DM automation covers the steps.
• Website. Paste the chat widget script before the closing body tag on your site. It works on any site builder. Details are on the AI chatbot for your website page.
• QR code. Print the QR code for your public Klugon page and put it in the window or on the bill folder.

Tools you probably already use, and how they fit

What a restaurant AI receptionist costs

Klugon is priced in US dollars, in three separate products. Restaurants usually start with voice, and add messaging if guests already message them.

Each phone number is $9 a month. A connected WhatsApp number is $19 a month. Extra voice minutes are $15 per started 100 minutes. Paying yearly takes 20% off. Full details are on the pricing page.

The examples below use assumptions we've made up for illustration. Replace them with your own.

Example 1: overflow and after-hours calls only

Assumptions:

• A neighbourhood restaurant forwards calls to Klugon only when nobody answers, and all calls after close.
• About 10 calls a day reach the AI.
• An average call lasts 2 minutes.
• 30 days in the month.

That's 10 × 2 × 30 = 600 minutes a month, inside the Starter allowance of 1,000.

Cost: $99 for voice Starter plus $9 for the number, so $108 a month, or $88 a month on yearly billing ($79 + $9).

Example 2: every call, plus WhatsApp

Assumptions:

• A busy restaurant sends every call to the AI.
• About 35 calls a day, with an average of 2.5 minutes each.
• Guests also message on WhatsApp, about 200 conversations a month.

Calls come to 35 × 2.5 × 30 = 2,625 minutes a month. On Starter that's 1,625 minutes over, which is 17 started blocks of 100 at $15, or $255 on top of $99. Business is cheaper at that volume: $299 with 3,600 minutes.

Cost: $299 voice Business + $9 number + $79 social Starter + $19 WhatsApp number = $406 a month.

Set against a person on the phone

From April 2026, the UK National Living Wage for workers aged 21 and over is £12.71 an hour.

Say you put someone on the phone and the door for 25 hours a week across Thursday to Saturday evenings and the lunch rush. That's 25 × £12.71 = £317.75 a week, or roughly £1,377 a month before employer's National Insurance, holiday pay and pension.

That's not a like-for-like comparison, and we wouldn't pretend it is. A good host greets people, manages the room, handles a complaint at the table and spots the regular who deserves a better table. An AI doesn't do any of that. The realistic comparison is whether the AI frees the host you already have from the phone during the hours when the door matters most. It also answers at 7am and midnight, when nobody is paid to be there.

What a missed booking is worth

Again, swap in your own figures. Suppose:

• 6 calls a week go unanswered during Friday and Saturday service.
• 1 in 3 of them would have booked.
• The average party is 3 people.
• Average spend is £30 a head.

That's 2 bookings × 3 people × £30 = £180 a week, or roughly £780 a month, walking to a competitor. If those numbers look anything like yours, the overflow set-up in Example 1 pays for itself. If you rarely miss a call, it may not, and you're better off spending the money elsewhere.

Our guide on how restaurants take reservations 24/7 goes further into the options, including ones that don't involve Klugon.

How to tell if it's working

After a month, look at four things in the dashboard:

1. Phone calls and minutes. How many calls the AI took, and when. The busiest-hours chart will show whether it's catching the Friday rush or mainly answering after close.
2. Bookings. How many tables were booked through the AI, by channel. Compare with the same weeks last year if you can.
3. Handled without a handover. The share of conversations the AI completed without flagging. If it's low, your information has gaps or your hand-over rules are too broad.
4. Questions without a good answer. The single most useful list. Each entry is something guests ask that the AI doesn't know. Fill those gaps every week for the first month.

Also ask your front-of-house team. If the phone at the host stand has stopped ringing during service and nobody is complaining about missed messages in the morning, that's the result you were after.

Cafes, takeaways and bakeries

Restaurants aren't the only food businesses with a phone problem, and the calls differ a lot by type. We've written separate guides for each:

• Cafes get fewer bookings and more "are you open?", "do you have oat milk?" and "can I bring the buggy?" questions. Many want a cafe AI assistant mainly for messages and the website, with the phone as a backup. The guide covers group bookings for brunch and catering enquiries.
• Takeaways live and die on the evening order rush. That guide is about AI takeaway ordering: what the AI can do with menu questions and order taking, the allergen rules for distance selling in detail, and how it sits alongside the delivery apps.
• Bakeries take custom cake orders weeks ahead, plus questions about what's left today. The bakery guide covers collecting cake order details (sizes, flavours, wording, dates) and handing them to the person who quotes.

If you run something in between, such as a bistro with a deli counter, read the guide closest to where most of your calls come from.

Frequently asked questions

Can an AI take restaurant reservations by phone?

Yes. Klugon answers the call, checks free slots in its built-in calendar using your opening hours and booking rules, books the table and gives the guest a reference on the call. It can't book into OpenTable, ResDiary or other reservation systems.

Does it send booking confirmations or reminders?

No. The guest gets their confirmation and reference in the conversation itself. Klugon doesn't send a separate email, text or WhatsApp confirmation, and it doesn't send reminders before the booking.

Can it take deposits or charge no-show fees?

No. Bookings made through Klugon don't include a payment step, so there are no deposits, card holds or no-show fees. If you rely on deposits, keep those bookings in your existing reservation system.

Can it transfer a caller to the manager?

No. It can't put a call through to a person. On a call it takes the caller's details and a note for your team, which appear on your leads list and in Slack or Zapier if connected, and tells the caller someone will call back.

How does it answer allergen questions?

Only from the allergen information you upload. Instruct it to list which dishes contain an allergen, never to call a dish safe, to tell the guest to speak to their server, and to pass the question to your team. Keep the matrix up to date whenever the menu changes.

Can it take takeaway orders?

It can answer menu questions and point callers to your online ordering page. It can also take an order down as a message for your team. It doesn't send orders to your till or kitchen printer and can't take payment on a call.

Can I use it only when we're busy or closed?

Yes. Set call forwarding on your existing line to forward only on no answer or busy, or after hours. Staff still pick up when they can, and the AI takes the rest.

Can I keep my restaurant's phone number?

Yes, by forwarding it. Klugon gives you a dedicated number for $9 a month and you forward your existing line to it with your phone provider. Number porting isn't supported.

Which languages can it speak on the phone?

English, German, French, Spanish, Italian, Portuguese, Dutch, Japanese and Arabic. You choose the language per assistant.

Can I try it before taking real calls?

Yes. The 7-day free trial lets you test the assistant in the browser with your own menu and rules. A card is needed to start the trial. Phone numbers can be added once the trial ends.

Klugon's AI receptionist won't run your restaurant, and it shouldn't try to. Set it up to take the bookings and questions that pull people off the floor, hand everything else to your team, and check the inbox after service.