Restaurant Waitlist Software: Where AI Helps at the Door

Partly. Waitlist software and AI phone assistants can quote times by party size, text guests when a table is ready and answer waitlist calls. A quote is only as good as the dwell-time data behind it, though, and a host is still needed for the kitchen backlog, large parties, access needs, fairness and recovering from a bad estimate.

The app roundups that rank for this topic compare features and prices. What decides whether a waitlist works is different: how does the app arrive at a number, and what does a person do when the number is wrong? TablesReady's page says wait times update automatically but not how they are calculated, and waitlist.me lists "Predictive Wait Estimates" without saying whether party size counts, so test any estimate against your own data before you rely on it.

Follow me on Instagram@sagnikteaches

Build your own quote table from POS data before trusting any app

A wait quote is a prediction of when a table of the right size will free up. The raw material is already in your point-of-sale system: when each check opened and closed, at which table, for how many guests. Export eight weeks of Friday and Saturday dinners, remove names and card details (you do not need them), and ask the AI for dwell times by party size:

Connect on LinkedInSagnik Bhattacharya

Attached is a POS export of checks for Friday and Saturday dinner for the last 8
weeks (check_id, table, party_size, opened_at, closed_at). Dwell time is
closed_at minus opened_at. Exclude checks under 20 minutes (voids) and over 3 hours
(likely errors) and list the ones you excluded. For party sizes 1-2, 3-4, 5-6 and
7+, give: number of checks, median dwell time and 80th percentile dwell time.
Also give the median gap between one check closing and the next one opening on
the same table.

An illustrative result for a 60-seat restaurant (ten two-tops, seven four-tops and two six-tops):

Subscribe on YouTube@sagnikteaches

The median gap between a table closing and the next check opening on it is 9 minutes, which is the real reset time, not the 5 minutes anyone would have guessed. Read the table for what it says about trust. The first three rows rest on dozens or hundreds of checks. The last rests on nine, so treat that row as a starting guess and let the host overrule it. If the AI gives you a precise-looking figure for a group of three checks, ask it to say so.

Two checks before you use the numbers. Pick five checks and work out their dwell times by hand, to confirm the AI used the right columns. And ask whether "closed at" is the time of payment, because guests who sit on after paying make your true turn time longer than the data shows. The 9-minute gap picks some of that up, since it measures from close to the next open. If your POS keeps no table field, the quote method below falls back to a single dwell time per party size and loses the table-by-table accuracy.

One quote worked through: a party of four at 7.15pm

With dwell times and the reset gap, a quote becomes a small queue calculation. A party of four arrives at 7.15pm on Saturday. The seven four-tops are all occupied, seated at 6.20, 6.35, 6.45, 6.55, 7.00, 7.05 and 7.10, and two parties that can use a four-top are already ahead of them. Using the 75-minute median for a party of three or four, the tables free up at 7.35, 7.50, 8.00, 8.10, 8.15, 8.20 and 8.25. Add the 9-minute gap and they can be reseated at 7.44, 7.59, 8.09 and so on.

The two parties ahead take the first two. Our party takes the third, at 8.09pm: a wait of 54 minutes. Repeat the sum with the 80th percentile dwell time of 90 minutes and the same table frees at 8.15, so the wait is about 69 minutes. The honest quote is therefore a range, rounded to the nearest five: 55 to 70 minutes. Now the host adds a judgement the data cannot hold: the kitchen is five tickets behind, so they say "about an hour, between 60 and 75 minutes". The app's default, in this illustration, would have told the party of four "35 minutes".

Ask the AI to write this as a rule you can paste into a staff guide or, if your app allows it, set as a quote formula:

Quote for a party = (time until the Nth table that fits them frees up, using the median
dwell time for that party size) + 9 minutes reset, where N = parties already waiting
for that table size + 1. Range = median to 80th percentile. Round to the nearest 5.
Host may add minutes for kitchen delay and must record the reason.

Most apps do not let you edit their formula, which is why the check in the accuracy log below matters. For a comparable method applied to another walk-in business, see AI for nail salons: walk-in wait estimates and technician matching, and for the data side, drafting restaurant pre-shift notes from POS data.

Reservations change the sum: tables you cannot offer to walk-ins

The 54-minute quote ignores the booking diary. In a restaurant that takes both reservations and walk-ins, most quote errors come from tables held for booked parties. Suppose two four-top reservations are due, one at 8.00 and one at 8.15. They take the third and fourth tables to free, so our party falls to the fifth, which frees at 8.15 and reseats at 8.24: a wait of 69 minutes, 15 more than before. With the 90-minute dwell time the range becomes about 70 to 85 minutes, and the host adds five for the kitchen.

To get the AI to do this, give it the floor and the diary as plain text:

It is 7.15pm. Seven four-tops were seated at 6.20, 6.35, 6.45, 6.55, 7.00, 7.05
and 7.10. Two walk-in parties are waiting for a four-top. Reservations for 4 at
8.00 and for 3 at 8.15 each need a four-top and have priority at their time.
Use a dwell time of 75 minutes (then repeat with 90) and a 9-minute reset.
When will a new party of four be seated? Show when each table frees up.

Check its table of free-up times against your own arithmetic once, then trust it for similar cases. What the AI cannot see is how a booked party behaves. A group that is 10 minutes late, or never arrives, is the host's call: many restaurants hold a booked table for 15 minutes past the booking time and then release it to the list, and the host decides when to enforce that. Whatever your rule is, write it down and put it in the quote prompt, because a table held for a no-show is a table the queue cannot use.

Texts that cut walk-outs without nagging guests

The text message is where waitlist software earns its fee. TablesReady's page lists a one-click text when a table is ready, with guests able to reply to confirm or cancel. The words matter more than the tool. Five messages cover a whole wait. Keep them plain: on most services a message of up to 160 plain characters is one segment, longer ones are split and billed per segment, and an emoji or curly punctuation cuts the limit to 70 characters.

1. Joined:   [Restaurant]: you are on the list for [4]. We estimate [55 to 70] minutes.
             We will text when your table is ready. Reply C to cancel.
2. Nearly:   [Restaurant]: nearly there. Please head back in the next 5 minutes.
             Reply C to cancel.
3. Ready:    [Restaurant]: your table for [4] is ready. We will hold it for 10 minutes.
             Reply C if you cannot make it.
4. Delay:    [Restaurant]: sorry, we are running behind. New estimate: [15] more
             minutes. Reply C to cancel.
5. Released: [Restaurant]: we have released your table after 10 minutes. Come to the
             host stand and we will do our best to seat you.

Three details to decide. The hold time (10 minutes here) should be the same in the text, on the sign at the door and in what the host says. The delay text should go out before the original estimate runs out, not after, so build it into the host's routine and not into the app. And message volume has a cost in some products: waitlist.me's pricing page, for example, includes 1,000 notifications a month on its entry plan and charges $0.02 for each one beyond that, so a busy room sending five texts per party should count them. Ask the AI to draft versions in the warm, short voice of your restaurant, then read each aloud. If a message sounds like a parking fine, rewrite it.

Phone calls: what Yelp Host and Slang do with a waitlist request

Many guests ask on the phone, "how long is the wait?", and the host is often busiest at exactly that moment. Two AI phone assistants are worth testing here. Yelp's product page says Host answers every call, even after hours, handles reservations and the waitlist in real time once your booking platform is connected, and lists automated waitlisting "with Yelp Guest Manager". When a caller needs a person, it transfers the call to a number you set in your Yelp for Business account. The page does not say whether Host reads out the current wait to a caller, so make that the first question in a demo, and if your waitlist runs on another app, ask whether Host can reach it at all. Decide in advance which calls it handles alone and which it passes to the host.

Slang's product page lists reservation integrations (OpenTable, Resy, SevenRooms and Yelp among them), routes VIP callers to staff and sends alerts for private-event enquiries, but it does not mention waitlists or wait times. Its pricing page lists Core from $399 and Premium from $599 per location without stating the billing period, and names no waitlist feature either. If waitlist handling matters to you, ask for a recorded demo of exactly that call. For a fuller view of each product, see whether Yelp Host is worth it for a small restaurant and Slang.ai for restaurants.

Whichever you test, write the call rules before the pilot:

The rule that matters most is that the assistant never quotes a time below the system's range. Test with ten calls of your own: a plain request, a noisy background, a party of eight, a caller who changes their mind, an angry caller, one who gives a number the assistant reads back wrongly. Listen to every recording for a week after launch. If you want a wider tour of phone options and costs, AI receptionist for restaurants: Yelp Host, Slang, Popmenu, Maple compares them.

What the host decides at the door, and what the log teaches

An app sees a list and a floor plan. A host sees the room. These are the calls that stay with a person, with the kind of moment that shows each one:

• The kitchen backlog. Tickets are 15 minutes behind, so every quote is short. The host adds minutes and records why.
• Access. A party of eight includes a wheelchair user, and the only workable setup is the corner six-top joined to a two-top. The host decides, and warns the floor.
• Who was first. Two parties both say they arrived before the other. The log settles the order, and the host settles the conversation.
• Incomplete parties. A party of four has three present. The policy says everyone must be there, and the host explains it kindly.
• Recovery. A guest has waited 25 minutes beyond the quote. The host offers a drink or a seat at the bar, within a budget the owner has set.
• Exceptions. An elderly guest, a child who has had enough, a regular with a booking that never showed. Each is a judgement about fairness that no formula makes.
• The table that lingers. A party that has paid and sat on for 25 minutes. Only someone on the floor can handle that gracefully.

The way to make the host's overrides useful is to record them. Give the host a one-tap reason when they change a quote or seat out of order: "kitchen", "access", "regular", "complaint". After two weeks, ask the AI to total the overrides by reason and by hour. You will probably find that "kitchen" accounts for most of the time added, and is the thing the quote formula should learn. A rule such as "add 5 minutes per 5 tickets over the usual" is better than a host remembering to do it, and it is a number the log gives you.

A two-week accuracy log: is the quote any good?

You can judge any app, or your own table, with a simple log. For two weeks, record for each party the party size, the quote they were given, the time they were seated, and whether they left before being seated. Forty parties is enough for a first read. The AI can total it in a minute. An illustrative result:

Overall, 22 of the 40 quotes, or 55%, landed within 10 minutes, and 7 parties left before they were seated, six of them parties of three or more. The pattern is clear: the app's quote is fine for couples and optimistic for anyone bigger. Until you can fix the formula, apply a manual offset by party size, plus 14 minutes for parties of three or four and plus 26 for five or six, and say it in the quote. Then rerun the log. If the share within 10 minutes rises from 55% towards 75% and the walk-outs fall, the change worked. If not, the cause is probably something else: a table plan the app does not know about, or turn times that vary more than the median suggests.

Keep your target modest and written down. "Most parties within 10 minutes" is checkable. "Accurate quotes" is not. And look at walk-outs alongside accuracy, since a very cautious quote can lose guests before they sit down, while an optimistic one loses them after they have waited too long.

A shift at the host stand with the tools in place

The routine below is short on purpose. Every step has a moment and an owner, which is what stops the system drifting back to "about half an hour for everyone".

1. Before doors open. Check that the quote table is current and the reservation holds are loaded. Make sure the text templates match the hold time on the sign, and that the phone assistant forwards to the right number.
2. At the door. Give the range and the reason: "55 to 70 minutes, the kitchen is a little behind". Offer the bar or a short walk. Take a mobile number and say which texts they will get.
3. Every half hour at peak. Compare the last three seated parties' actual waits with their quotes. If all three ran over by 10 minutes or more, add 5 minutes to every new quote and send the delay text to the parties already waiting.
4. When a table is ready. Send the ready text and hold the table for the stated 10 minutes. After that, send the released text and give the table to the next party that fits.
5. After close. Spend ten minutes recording walk-outs and overrides, each with a reason. Those two lists feed the next fortnight's log.

If your software offers a self check-in kiosk, as waitlist.me does on its entry plan, it takes some of the first step off the host. It also invites wrong party sizes, since guests type their own, so the host still glances at the list against the people at the door. Once the routine runs, the AI's work is on the evenings after: totalling the logs, spotting the party sizes the quote misses, and drafting the next change for you to approve.

What the apps cost and what each one leaves to you

Prices below are list prices from each vendor's pages in October 2026, in USD, and they change, so check the live page. I have left out features I could not confirm.

For a small room, the decision usually comes down to one question: does the software you already use for reservations have a waitlist, and does it send texts? If it does, test its quotes with the log before you pay for anything else. For a wider comparison of booking tools, see reservation tools compared for small restaurants. Add an AI phone assistant only if calls at the busiest hour are an actual problem: count them for a week first, and listen for how many are wait-time questions.

The workflow that works is deliberately plain. Build the dwell-time table from your POS, apply it as the quote, let the app send the texts, let the phone assistant take the routine calls, and keep a host at the door making the calls no app can. Review the override log and the accuracy log every two weeks, and change one thing at a time. When the host says the quotes feel right, check the numbers, because that is the moment the log is most useful.

Questions owners ask about AI at the host stand

Can AI take over the host stand completely?

Not safely. It handles the routine parts well: calculating a quote, sending texts and answering calls when the host is busy. It cannot see a kitchen that is falling behind, a guest who needs step-free access or an argument about who arrived first. Keep a person at the door at peak times and let the software support them.

How accurate should a wait quote be?

Set your own target and measure it. A reasonable aim is for most parties to be seated within 10 minutes of the quote, with a wider margin for large parties. Quote a range rather than a single time, and track the gap between quoted and actual wait for two weeks before you change anything.

Should I quote a wait when I am not sure?

Yes, as a range, with the reason: for example, 55 to 70 minutes because the kitchen is behind. Guests cope better with an honest range than with a confident number that slips. If the estimate changes, text the new one before the old one has run out.

Can I text guests who walked up to the stand?

Ask for consent when you take their number, say which messages they will get and how to stop them, and use the number only for that waitlist. Keep marketing messages separate and ask your adviser about the consent rules where you operate.

Further reads

• AI Chatbot for Restaurants: Hours, Bookings and Hand-Offs to Staff — How a chat assistant should hand over to staff.
• AI Bot for Restaurant Reservations: Chat Booking Setup and Cost — Set up chat booking and see what it costs.
• AI Phone Answering for Restaurants: What It Costs and Pays Back — What phone answering costs and how it pays back.
• AI Receptionist for Barbers: Walk-In Queues, Waitlists and Texts — Walk-in queues and text updates handled in another trade.
• Can AI Answer the Phone and Take Restaurant Reservations? — Whether AI can take reservations by phone at all.
• How to Build a Restaurant Staff Rota With AI Demand Forecasts — Staff the host stand from the same demand forecast.
• AI Software for Restaurants: POS-Built AI or an Add-On Tool? — Compare Toast IQ, Square AI and Lightspeed AI with add-on tools, job by job, with list prices and a checked date.
• Restaurant No-Shows: Deposit Rules by Party Size, AI Confirmations — Set deposit and card-hold rules from your own no-show log by party size, then use AI for policy wording, reminder texts, reply sorting and waitlist offers.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: TablesReady home and pricing pages, the waitlist.me pricing page, Slang's product and pricing pages, and Yelp Host's product page, all as of October 2026. Dwell times, quotes and results in the examples are illustrative.