SevenRooms Voice AI answers every guest call, checks availability, books and changes reservations, and writes the call into the guest's profile. It is an add-on to SevenRooms' own packages, sold by custom quote only, with no published price. It suits busy venues already on SevenRooms; others should compare cheaper tools first.
The catch is the order of decisions. Voice AI is a tool for a restaurant that has already chosen SevenRooms as its booking and guest-data system. If you have not, the real question is which phone assistant fits the system you do use, and this tutorial treats Voice AI as the premium end of that comparison.
What it actually does on a call
SevenRooms describes Voice AI as answering every guest call around the clock, booking reservations and giving personalised responses tailored to each venue and caller. Its product page lists new bookings, changes and cancellations by voice, answers on hours, parking, menus and policies, greetings for returning guests based on their profile, deposits and payment links at the time of booking, transfer of complex requests to a designated staff number, and call recordings in a dashboard. Ask SevenRooms to demonstrate each one on a call to your own number.
Here is a sample call. The words are illustrative, not a transcript:
Caller: "Hi, do you have a table for four on Friday around eight?"
AI: "I can offer 7:45 or 8:30 on Friday for four. Which would you prefer?"
Caller: "8:30. It's a birthday, can we have a window table?"
AI: "Lovely. I've noted a birthday and a window request, though I can't
guarantee the table. Can I take a name and mobile number for the
confirmation text?"Two things in that exchange matter more than the voice. The AI read live availability, and it wrote down a request without promising it. If the demo does either badly, move on.
The part that justifies the price: guest records
SevenRooms says Voice AI draws on the restaurant's CRM and guest profiles, recognises returning guests from their reservation history and sends every call and booking to its reporting dashboard, where it also tracks the booking revenue the assistant generated. If you already use the CRM for birthdays, allergies and regulars, a phone assistant that updates it is worth more than one that only fills the diary. If you do not use the CRM, you are paying for a feature you will not use.
Which plans include it
None of them, by default. SevenRooms (now a DoorDash company) lists three packages on its pricing page, Starter, Growth and Premium, and a set of add-ons that includes Voice AI beside email marketing, event management, WhatsApp and text marketing, integrations and a group portal. Each package carries a "get a quote" button rather than a price. So the honest answer to "which plans include it" is that it is a separately quoted add-on on top of whichever package you buy, and the price depends on your package, venue count and call volume.
That is why figures that circulate online are unreliable. Monthly prices quoted on review sites are estimates, and some come from competitors. Use them to sense-check a quote, never to build a budget.
What to ask for in the quote
A custom quote is negotiable only if you know what to ask. Send this list before the sales call, and ask for answers in writing.
1. Price: monthly fee for Voice AI on top of my package, per venue.
2. Usage: any call cap, per-minute charge or overage? What counts as a call?
3. Term: monthly or annual? Cancellation notice? Early-exit fee?
4. Set-up: one-off fee? Who writes the knowledge (hours, menu, policies)?
5. Pilot: can I trial it for four weeks with recordings and a clean exit?
6. Transfer: what happens when the AI cannot help? Does it ring a staff phone?
7. Deposits: can it take card holds or send payment links, and how are disputes handled?
8. Data: who owns the call recordings and transcripts? How long are they kept?
9. Languages: which are supported, and how do I check the quality?
10. Reporting: do I see calls answered, booked, transferred and dropped?Questions 2 and 6 matter most. Usage terms decide whether a busy Friday costs more, and the transfer route decides whether a guest with an awkward request gets a person or a dead end.
The cheaper rivals, and what they publish
A comparison is only fair if each tool's pricing is stated honestly, including where it is unknown.
| Tool | Published price | Notes |
|---|---|---|
| SevenRooms Voice AI | None; quote only | Add-on to SevenRooms packages; writes calls to guest profiles |
| Yelp Host | From $249 a month after a free trial of up to 30 days (Yelp's own table shows $249 to $399); bundled savings with Yelp Guest Manager | Pulls hours and details from your Yelp page; transfers calls it cannot handle to a number you set |
| Popmenu AI phone answering | Its product page says from $149 a month, with custom quotes | Texts reservation or ordering links during the call; ask whether it needs a Popmenu platform plan |
| Slang | No price on its own pricing page; Core and Premium tiers plus a higher custom tier | Integrates with OpenTable, SevenRooms, Yelp and Resy; private events capture on Premium |
| Maple Voice | $150 a month; Pro $350 | Phone ordering as well as bookings |
Slang's tiers matter here because it connects to SevenRooms. A SevenRooms venue that wants phone AI is not forced to buy Voice AI; it can compare Slang's integration with SevenRooms' own add-on. Competitor comparison pages quote Slang prices per location, but Slang's own page shows none, so ask Slang directly and get the quote in writing.
Yelp Host is the benchmark for a published price. Its starting price rose to $249 a month by October 2026, so older reviews quoting a lower figure are out of date. It transfers calls it cannot answer to a number set in your Yelp for Business account, so test what happens when that phone rings unanswered during service. For a fuller look at the cheap end, the Yelp Host tutorial weighs it for a small restaurant, and the Slang tutorial covers the main rival.
Who Voice AI suits
A busy venue already on SevenRooms
If the restaurant uses SevenRooms for reservations, waitlist and guest notes, the add-on keeps everything in one place. There is no second system to reconcile, and a regular's phone booking lands in the same profile as their online one. This is the clearest fit.
A group with several venues
Groups benefit from one guest record across sites and one reporting view. They are also the type of buyer who can negotiate a quote, because the volume is real. Ask about multi-venue pricing and whether each venue's knowledge can differ.
A restaurant with high-value tables and high phone volume
A fine-dining room where the average party spends a lot, and where guests ring to ask about tasting menus and dietary needs, gains more from a considered assistant than a pizza takeaway does. The value sits in the information the AI captures, so check how it records allergies and notifies the kitchen.
Who should look elsewhere
An independent on another booking system, a cafe whose calls are mostly "are you open?", or a takeaway whose calls are orders would all do better with a tool built for that job. For takeaway, the takeaway phone line tutorial is a better starting point.
Eight call types, and how each should be handled
Whatever you buy, write down what should happen for each type of call. This table doubles as your demo script and your pilot scorecard. Vendors rarely volunteer the awkward rows, so ask for them.
| Call type | What good looks like | Test it by asking |
|---|---|---|
| Standard table for two to four | Reads live availability, books, sends a confirmation text | A table for 3 tomorrow at 7:30 |
| Large party (say 8 or more) | Collects details and passes to a manager; does not book on its own | A table for 12 on Saturday |
| Allergy or dietary need | Records the request exactly, flags the booking, never says a dish is safe | "Is the tasting menu safe for a severe nut allergy?" |
| Private dining or events | Captures date, numbers, budget and contact, then hands to the events lead | A 40-person birthday in March |
| Change or cancellation | Finds the booking by name and phone number, changes it, confirms by text | "Can I move tonight to 9?" |
| Complaint or refund | Takes a message, apologises once, promises a call-back time | "The food last night made me ill" |
| Lost property or a delivery driver | Takes a message for the right person, does not try to solve it | "I left a scarf on Friday" |
| Sales calls and wrong numbers | Ends politely without logging a booking | A supplier pitch |
Score each row from 0 to 2 during the demo. A tool that is excellent on the first row and shaky on the allergy row is a risk, however natural it sounds. A hospitality group can learn a lot from a single demo call about a severe allergy, because the right response is not a clever answer but a recorded request and a hand-off.
A second example: a private-events enquiry at 11pm
Illustrative: a caller rings after closing to ask about a 30-person birthday dinner next month. With voicemail, the restaurant gets a muffled number and a missed lead, and the caller books somewhere that answered. With an assistant that captures the enquiry, the owner opens a message in the morning.
Event enquiry captured overnight (illustrative)
Name: [first name], mobile ending 4821
Event: 30th birthday dinner, 30 guests, Saturday 14 March
Needs: private area if possible, vegetarian menu for 6, cake allowed
Budget: not given
Asked: whether there is a minimum spend
Status: not booked, owner to call back by 11am
The value is not the AI's voice. It is that every enquiry arrives in the same structured shape, ready for a human to price. A restaurant that gets four events enquiries a month, converts one in four and earns $1,500 on each event is at about $1,500 a month from those calls. If the assistant lifts the capture rate from half of those calls to nearly all of them, the extra enquiries matter more than any booking-diary gain. Check whether events leads go into the same system as events management and what the quote says about that module.
Tell guests it is an assistant, and tell staff what changed
A short greeting that names the restaurant and says it is an automated assistant avoids the awkward moment when a guest realises mid-call. Something like this works:
"Thank you for calling [restaurant name]. I'm the restaurant's automated
assistant, and I can book a table, change a booking or answer questions.
If you'd rather speak to someone, just say so."Brief the team too. Staff need to know three things: where the AI's bookings appear, how to see a transcript when a guest says "I spoke to someone yesterday", and which calls arrive on their mobile. A phone assistant that surprises the floor team becomes a source of grumbles within a week. A five-minute pre-service briefing for the first fortnight, and a shared note of anything the AI got wrong, is enough.
Choosing between Voice AI and the alternatives, by situation
You are on SevenRooms and use guest notes. Request the Voice AI quote and a pilot, and ask Slang for a comparison quote, because Slang integrates with SevenRooms and may cost less. Pick on the pilot scorecard, not the demo.
You are on SevenRooms but ignore its CRM. The premium case weakens. A plain phone assistant that books into the diary is enough, so compare Slang and Yelp Host first.
You are on another reservation system. Voice AI is probably not for you. Check which assistants support your system and ask for the integration to be shown live.
You want phone orders as well as bookings. Look at tools that handle takeaway orders and modifiers; the four-way receptionist comparison covers who does what.
You are a small independent with limited budget. Start with the lowest published price, run a 30-day trial where one is offered, and keep the call log to decide. You can always step up to a premium add-on once you know your missed-call number.
A worked break-even example
Illustrative figures for a 60-seat restaurant on SevenRooms. It receives about 90 calls a week, and staff miss around 25 of them during service. Suppose a third of missed callers would have booked a table of three on average, and each person spends $35. Then:
Missed calls a week 25
Would have booked (one third) about 8
Covers per booking 3
Average spend per cover $35
Weekly revenue recovered 8 x 3 x $35 = $840
Realistic recovery (60%) about $500 a week, about $2,150 a month
Gross margin on that revenue take 60% = about $1,300 a monthIf the quote is $400 a month all in, the add-on clears about $900 a month on these assumptions. If the quote is $1,200, the result is much thinner. Swap in your own figures, and count missed calls from your phone log for two weeks rather than guessing. The margin figure is your food and drink cost plus any extra staffing, which only you know.
The same sum applies to every phone assistant, which is why the price per call matters less than the number of calls you actually miss. The sums in the phone answering cost tutorial use a similar method.
A quick sum on the quote itself
Suppose the quote arrives as a $350 monthly fee plus a one-off set-up charge of $500, on a twelve-month term. The first-year cost is 12 x $350 + $500 = $4,700, or about $392 a month. Divide that by the number of extra bookings you expect. If you expect 20 extra phone bookings a month at an average of $90 a booking, that is $1,800 of revenue and the add-on costs about 22% of it before food and labour. Run the same sum with a pessimistic figure of 8 extra bookings. If the answer still works, sign. If it only works on the optimistic figure, negotiate a shorter term or a pilot.
A pilot plan you can run in four weeks
- Week 0. Log every call for seven days: answered, missed, what the caller wanted. This is your baseline.
- Week 1. Write the knowledge: opening hours, menu notes, allergens, parking, large-party and private-room rules, deposit policy. Test five calls yourself, including one about a nut allergy.
- Weeks 2 and 3. Run it live. Listen to ten recordings a week. Mark each as correct, clumsy or wrong.
- Week 4. Compare booked covers from phone calls with the baseline, count transfers and complaints, and ask staff how often they had to fix a booking.
Set a pass mark before you start. For example: at least nine out of ten calls handled without a staff fix, no booking made for a wrong date, and no allergy question answered without a staff note. If the pilot falls short, you have evidence for a cheaper tool or a better quote.
A realistic failure to plan for
Illustrative: a guest rings about a wheelchair-accessible table. The assistant confirms a table but does not know which tables are step-free, because the knowledge base never said. The fix is to list accessibility facts explicitly (step-free entrance, accessible toilet, table numbers that suit) and add "any access question not in this list is transferred to a person". This type of gap rarely shows in a demo, so find it in the pilot.
How to review the recordings without losing an evening
The pilot depends on listening, and listening takes time unless you do it in a fixed routine. Block 30 minutes on the same day each week, pull ten recordings (five from busy periods, five from quiet ones, and always any call that ended in a transfer), and mark each on a one-line sheet.
Date | Call type | Booked right? | Wrong detail? | Tone ok? | Fix needed
Fri | table of 4 | yes | no | yes | none
Fri | allergy | n/a | said "should be fine" | no | add rule: never confirm safety
Sat | change | yes | no | yes | none
Sun | event | partial | missed budget | yes | add budget questionThe "fix needed" column becomes your list of changes to the knowledge base. After two rounds most owners find the same three or four gaps closed, and the failure rate drops. If a vendor does not give you recordings and transcripts at all, that alone is a reason to walk away, because you cannot fix what you cannot hear.
One more habit: ring the restaurant yourself on a Sunday evening once a month, as a guest, and ask for something slightly unusual. It takes five minutes and finds drift, such as an outdated opening time left in the knowledge base after a seasonal change.
Risks to name in the contract
Call recordings are personal data, so ask for the retention period and a way to delete a guest's recording. Agree who answers a dispute when the AI books the wrong date. Check that the voice announces itself as automated if the rules where your guests live require that, and ask a qualified adviser if you are unsure. Make sure you can take your guest data with you on exit, because a guest database is the part of the system that gets harder to move over time.
For a wider checklist, the restaurant AI vendor questions apply to any phone-AI contract.
The decision in one pass
If you run SevenRooms and use its guest records, ask for a Voice AI quote and a four-week pilot. If you run another system, start with Slang, Popmenu or Yelp Host according to the stack you have, and spend the saving on a better knowledge base. If you only miss a few calls a week, answer the phone yourself on Friday and Saturday nights and revisit in a month. The cheapest phone assistant is the one you do not need.
Questions to settle before you ask for the quote
Does SevenRooms Voice AI work without the SevenRooms booking system?
SevenRooms presents it as an add-on to its own packages, so assume you need to be a SevenRooms customer. If you run another reservation system, compare tools built to plug into several systems, such as Slang, and ask each vendor for a written list of supported integrations.
Is there a published price for SevenRooms Voice AI?
No. SevenRooms' pricing page lists Starter, Growth and Premium packages and add-ons, and each one says get a quote. Figures you see elsewhere are third-party estimates or competitor claims, so treat them as a prompt for questions, not a budget.
Can the AI take a payment or deposit over the phone?
SevenRooms' Voice AI page lists prepaid booking support: collecting deposits and sending secure payment links at the time of booking. Ask how that maps to your own deposit and no-show rules, whether card holds work the same way, and who handles a guest who disputes a charge.
What happens when the AI cannot answer?
SevenRooms says complex requests and upset guests are transferred to a designated staff number in real time. Ask what happens when nobody picks up that number during service, and test it with a live call before signing.
How long should I trial it before deciding?
Ask for a pilot of at least four weeks across a quiet week and a busy weekend, with call recordings. Count missed and booked calls before and after. Anything shorter tests the novelty, not the service.
Further reads
- Popmenu AI Answering: Does It Take Orders or Only Send Links? — Whether Popmenu takes orders or only sends links.
- Maple AI for Restaurants: Voice or Pro Plan, and Who Needs Which — Maple Voice plans for bookings and phone orders.
- Restaurant AI Voice Agent: 12 Test Calls to Run Before Go-Live — Twelve test calls to run on any voice agent.
- Restaurant No-Shows: Deposit Rules by Party Size, AI Confirmations — Deposit rules and confirmations that cut no-shows.
- Can AI Answer the Phone and Take Restaurant Reservations? — Can AI really take a reservation by phone?
- AI Startups for Restaurants: Protect Your Number, Data and Exit — Exit terms to agree before signing any AI vendor.
- How Much Does AI Cost a Small Restaurant Each Month? — The wider monthly bill for AI in a small restaurant.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: SevenRooms pricing and Voice AI pages, Yelp Host product page, Popmenu AI answering page, Slang pricing page, all checked October 2026.