AI Phone Answering for Restaurants: What It Costs and Pays Back

Restaurant AI phone answering lists at about $85 to $600 a month per location (September 2026): Maple from $85 billed yearly for answering only, Slang.ai from $399 with reservations. Payback is quick if you miss calls: at $45 a head and a 65% margin, about five extra tables of two a month covers a $300 plan.

The subscription is only the visible part. Setup time, integrations that sit on a higher tier, caps on callers and your phone provider's forwarding charges decide the real bill. Payback depends on how many calls you miss today, which most restaurants have never counted. Both halves are below: the full cost, and the maths to run on your own numbers.

Follow me on Instagram@sagnikteaches

Published prices, September 2026

These are list prices per location from the vendors' own pages. Several restaurant-specific vendors don't publish prices at all, so treat the table as a guide to the range, not a full market.

Connect on LinkedInSagnik Bhattacharya

You can check current figures on Maple's pricing page and Slang.ai's pricing page. The main divide in every price list is the same: answering questions is cheap, and writing into your reservation book or POS costs roughly double. Whether an AI can take bookings properly, rather than just messages, is covered in whether AI can answer the phone and take reservations.

Subscribe on YouTube@codingliquids

Costs the pricing page leaves out

The knowledge-base hours go further if you write each answer the way you'd want it said aloud, with the limits spelled out. A before and after from an illustrative tapas bar shows the difference:

BEFORE
Groups: We love hosting groups! Just give us a call.
Dietary: We can cater for most dietary requirements.

AFTER
Groups: We take bookings for up to 8 online or by phone.
For 9 to 16 people, take the name, date, time and number,
say the manager will call back within one working day to
confirm, and do not promise a table. A $10 a head deposit
applies to groups of 9 or more.
Dietary: Say we have vegetarian and vegan dishes marked on
the menu. For any allergy or intolerance, do not say a dish
is safe. Take the caller's number and pass it to the
manager on duty, who will check with the kitchen.

The "before" lines read well on a website, but read aloud by an assistant they become promises: "just give us a call" invites it to book a party of 14 on a Saturday, and "most dietary requirements" invites it to reassure someone with a nut allergy. Every line that describes a limit, a number or a hand-off to a person is a line the assistant can follow.

Whether it does follow them shows up in the call summaries, so read every one for the first two weeks. An illustrative summary from the same tapas bar's first Friday:

Call summary - Fri 18:42, 2 min 10 s
Caller:   [first name], mobile ending 4471
Intent:   Reservation
Details:  Saturday 20:00, party of 11
Outcome:  Table booked, confirmation text sent
Flags:    none

That booking should never have happened: 11 is above the limit of 8, so the right outcome was a callback request for the manager, and "Flags: none" means the hand-off rule didn't fire at all. The fix took two steps. The group rule moved to the top of the knowledge base as a hard instruction, and the bar asked the vendor whether bookings made by the assistant respect the maximum party size set in the reservation system; where they do, setting the cap there means the limit holds even when the instructions slip. The cost of missing it is real: a Saturday table for 11 the kitchen hadn't planned for, or a phone call to un-book a guest who thought it was settled.

Two items in that table catch restaurants most often. Answering-only plans sound cheap until you realise they can't put a booking in the diary, so every booking call becomes a message someone must return. And per-caller pricing, which looks generous for a dentist or plumber, fills up fast for a restaurant: every person asking whether you have a table on Saturday is a unique caller.

Count your missed calls before you price anything

Payback depends almost entirely on one number: how many calls go unanswered now, and what those callers wanted. Two weeks of counting is enough.

1. Pull the call log. Most business phone systems and many mobile carriers show missed and unanswered calls with times. If yours doesn't, put a tally sheet by the phone.
2. Sort every call into four types: booking attempt, takeaway or collection order, simple question (hours, parking, menu, dogs, dress code), and everything else (complaints, suppliers, lost property).
3. Note the time of each missed call. Expect them to cluster in service peaks, when whoever answers the phone is also seating guests or running food.
4. Check how many missed callers called back. A missed call from a number that rang again ten minutes later and booked isn't a lost booking.

A filled-in count can end the discussion quickly. Here is two weeks at an illustrative 36-seat café-bistro that serves breakfast and lunch daily and dinner on Fridays and Saturdays:

About 47 missed calls in a fortnight sounds like a problem, but only four were lost business, roughly eight a month, and two-thirds of all missed calls were people asking about opening times, the high chairs and whether dogs are allowed. The case here is about staff being interrupted during the lunch rush, not lost revenue, which points to an answering-only plan or simply a better website and Google profile. Compare that with the restaurant below, where the same count points the other way.

This count also becomes your baseline for judging the tool later, which is exactly what measuring an AI receptionist's return over 90 days relies on.

Payback maths with your own numbers

Lost booking calls per month      (missed, and never called back)
x share the AI would convert       (assume 70-80%, then check)
x average covers per booking
x average spend per cover
x contribution margin              (after food and drink costs)
= recovered contribution from bookings

+ the same calculation for takeaway orders
+ staff time freed x hourly cost   (count it only if you use it)
- monthly plan cost, overages and your update time
= monthly payback

Here's an illustration. Say a 55-cover neighbourhood restaurant, open six nights, takes about 160 calls a week and misses 28 of them, mostly between 6 and 8pm. Its two-week count shows 10 missed booking attempts a week, 4 missed takeaway orders and 14 questions. About half the booking and order callers never ring back.

• Bookings: 5 lost a week, of which the AI converts 80%, so 4 a week or about 17 a month. At 2.6 covers per booking and $45 a head, that's about $1,990 of revenue. At a 65% contribution margin after food and drink costs, roughly $1,290.
• Takeaway: 2 lost orders a week, 80% converted, about 7 a month at $38 each: around $270 of revenue and $170 of contribution.
• Cost: it needs bookings and orders, so an answering-only plan won't do. Maple Pro at $350 a month billed monthly, plus about two hours a month of updates.

That's around $1,460 of recovered contribution against roughly $400 of cost: comfortably positive. Now the cautious version. If only 2 booking callers a week are truly lost, recovered contribution from bookings falls to about $520 a month, which still covers the plan, but only just once you count setup time. And if the count shows you miss five calls a week and most are questions, the case is about service and staff interruptions, not money; an answering-only plan at $85 to $150 is the most you should spend.

Three budgets: answering only, bookings, orders and payments

First-year totals below include about 8 hours of setup and 30 minutes a week of updates at $25 an hour, around $850 of staff time. Your own rate will change the totals, so treat them as illustrations.

Multiply by locations if you have more than one; every vendor above prices per location. If takeaway is most of your phone volume, the setup questions are different, and setting up an AI phone line for takeaway orders goes through them.

Reading a vendor's savings claim

Vendors publish headline savings, and some are specific. Kea's homepage, for instance, says restaurants save $677 a month on average in missed order revenue. A figure like that is an average across the vendor's own customers, who chose the product because they had a phone problem, so it tells you little about your restaurant. Three questions turn a claim into something you can use:

• Revenue or margin? "Missed order revenue" is the till total. Your gain is the contribution after food and packaging costs, which is well under the headline. Take the $677 figure at face value and assume, for illustration, that food and packaging are 35% of a takeaway order: the contribution is about $440 a month. Take off a $300 plan and the net gain is closer to $140 than $677, before any setup time.
• Compared with what? Ask whether the baseline was measured before the product went live, or estimated afterwards from calls the AI answered. Calls it answered aren't all calls you'd otherwise have lost.
• For a restaurant like yours? Ask for the figure for customers with similar covers, opening hours and takeaway share. If the vendor can't break it down, rely on your own count.

Where the payback case falls apart

• You don't miss many calls. If staff answer 95% of calls, the AI is recovering very little. The count settles this.
• The AI can't reach your diary. Without a reservation or POS integration, it takes messages, and callbacks eat the time you hoped to save.
• The menu is out of date. An assistant that confidently offers a dish you took off last week costs more goodwill than a missed call.
• Allergen questions. These should reach a person who can check with the kitchen, not a script. Whether AI should answer allergen questions explains how to route them safely.
• Your guests want a person. A small, regular-heavy restaurant may find the AI better as an overflow than as the first voice callers hear.

That last point often shows up as a setup slip rather than a policy choice. An illustrative family-run trattoria meant to send only unanswered calls to the AI, but the forwarding was set to "all calls" during setup. For the first week, regulars who had rung the same number for years heard an automated voice even when the owner was standing next to the phone. Three of them mentioned it in person, and one left a review about it. Nothing was wrong with the assistant; the fix was changing the forwarding to "no answer after four rings, or busy" in the phone provider's settings, and checking it by ringing the restaurant from a mobile at a quiet moment.

If a person on the phone matters more to you than the cost, compare with a human service; AI receptionist or answering service sets the two side by side.

Ways to cut the bill without losing the benefit

• Forward only unanswered calls. Staff still pick up when they can; the AI catches the rest, which is where the value is anyway.
• Start monthly, switch to yearly later. Pay the higher monthly price for 60 to 90 days, prove the recovered bookings, then take the annual discount. On Maple Voice, three trial months at $150 cost $450, against $255 for the same three months billed yearly, so the flexibility costs about $195. That's cheap insurance against a year-long contract for a tool that turns out not to fit, and cheaper still if your count was borderline.
• Buy the tier your missed-call mix needs. Mostly questions: answering only. Mostly bookings: the reservation tier. Don't pay for payments and upselling you won't use.
• Ask for the setup fee to be waived, especially if you commit annually after a trial.
• Book the review now. Put a date 90 days out to compare missed calls, recovered bookings and cost, and cancel if the numbers aren't there.

Questions restaurant owners ask about AI phone lines

Can I keep my existing restaurant phone number?

Usually yes. The common setup forwards calls from your existing number to a number the AI service provides, either every call or only those that ring out or come in while the line is busy. Check which forwarding options your phone provider offers and whether it charges for forwarded minutes. Moving (porting) your number to the AI vendor is rarely necessary and makes leaving harder.

Will regular customers be put off by an AI answering?

Some will. Reduce it by forwarding only unanswered calls, so staff still pick up when they can, and by giving callers a clear way to reach a person. Have the assistant greet callers with the restaurant's name and say it's an automated assistant. Read call summaries and complaints closely in the first month; that tells you more about your own guests than any vendor claim.

Do I need the POS or reservation integration from day one?

Not necessarily. If most of your missed calls are questions about hours, parking or menus, an answering-only plan handles them. If most are booking attempts, an answering-only plan just takes messages, and someone still has to call back, which is the work you were trying to remove. Count your missed calls by type first; the mix tells you which tier you need.

Further reads

• Restaurant AI Tools: 12 Questions to Ask Before You Sign Up — The twelve questions to put to any vendor before you sign.
• How Much Does an AI Receptionist Cost for a Small Business? — How general AI receptionist pricing works outside hospitality.
• What to Do When an AI Receptionist Gets a Booking Wrong — What to do when the AI books a table wrongly.
• How to Train Front-of-House Staff to Work Alongside AI — Get hosts and servers working alongside the new phone line.
• AI Call Summaries: Log Every Phone Enquiry in Your CRM — Turn every AI-handled call into a record you can act on.
• What an AI Implementation Plan Looks Like for a Small Restaurant — Where a phone line fits in a wider restaurant AI plan.
• How Much Does AI Cost a Small Restaurant Each Month? — A line-by-line monthly AI budget for a small restaurant, three priced set-ups, and the staff hours that cost as much as the subscriptions.
• AI Ordering vs Delivery Apps: What a Takeaway Actually Pays — Delivery app commission against AI phone and web ordering, line by line, with a break-even worksheet a takeaway can fill in with its own numbers.
• Where AI Actually Saves Time in a Small Restaurant — A task-by-task look at a small restaurant's admin week: where AI cuts real minutes, where it only moves them, and where it adds work.
• 7 AI Mistakes Restaurant Owners Make With Bookings and Reviews — Seven ways restaurant AI goes wrong with tables and reviews, each with a real-looking example, what it costs and how to fix it.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Maple, Slang.ai and Goodcall pricing pages; Loman and Kea product pages. Checked September 2026. Payback figures are illustrative assumptions, not vendor claims.