# AI for restaurants: 5 workflows that cut no-shows, fill seats, and stop losing money on empty tables
The short version: A restaurant running OpenTable or Resy, Toast POS, and 7shifts can automate five workflows today. Cut no-show rates from 20-25% down to 6-8%, respond to every Google and Yelp review within the hour, catch schedule gaps before they become understaffed Friday nights, and catch vendor billing errors you'd otherwise never find. A single-workflow build runs $2,495-$4,995 fixed, live in 7 days. Here's what each one actually looks like.
The empty table problem
A Saturday at 7pm. You're fully booked on paper. Twelve tables are reserved. By 7:15, three parties haven't shown. You turned away a walk-in group of four at 7:05 because you "didn't have room." You now have room for eight people and none of them are there.
This happens 20-25% of the time without reminders. Hospitality industry benchmarks put the no-show rate at 20-25% across reservation platforms without proactive outreach. With structured reminders (not just the automated OpenTable email that everyone ignores), that rate drops to 6-8%. The table you lost to no-shows on a 60-seat restaurant at $60 average check is roughly $3,600 in revenue per weekend night, assuming 20% no-show rate. Not theoretical. Gone.
The five workflows below address that and the four other places where restaurants lose money they've already earned.
The 5 workflows worth automating first
1. No-show reduction through multi-touch reminders
What happens now: OpenTable or Resy sends one automated confirmation email when the reservation is made. Some restaurants send a reminder 24 hours out. Most rely on that one email and hope for the best. The email confirmation open rate for restaurant reservations sits around 42%. That means more than half your reserved guests haven't explicitly confirmed they're coming.
What the agent does: 48 hours before the reservation, a text goes out to the guest's mobile number from your restaurant's business line: "Looking forward to seeing you Thursday at 7:30, [Name]. Your table for 4 is confirmed at [Restaurant]. Reply YES to confirm or call us if your plans change and we can help someone else get that table." 24 hours out, if no reply, a second message with a lighter tone. 2 hours before the reservation, a final reminder with your address and a link to your parking info.
Guests who reply NO or cancel get immediate follow-up: "No problem at all. Should I add you to our cancellation list for another night?" The table goes back into the available pool in OpenTable or Resy automatically, and a cancellation notification fires to your front-of-house.
Real number: Industry data from SevenRooms and Resy's own published case studies puts average no-show reduction at 60-70% with a structured multi-touch SMS reminder flow. Moving from 22% no-show to 8% on a 60-seat restaurant doing 3 seatings on a Saturday night at $58 average check is roughly $1,800 recovered per Saturday. That's not the year-end tally; that's one night.
2. Google and Yelp review response
What happens now: Reviews come in while you're in service. You see them at 11pm when you're closing or on Sunday morning when you're off. The 3-star review that misunderstood your no-substitutions policy on the prix fixe sits unanswered for four days. Google's algorithm factors recency and owner response rate into local search ranking. An unanswered 3-star review is doing more damage than the rating itself.
What the agent does: When a new review posts to Google or Yelp (via webhook or polling), the agent reads the content and drafts a response within 15 minutes. For 5-star reviews, the response is specific to what the guest mentioned ("So glad you loved the duck. It's been on the menu since we opened in 2019 and the kitchen would be happy to hear that"). For 3-4 star reviews, the agent drafts a response that acknowledges the specific issue, explains context where relevant, and invites the guest back. For 1-2 star reviews, the agent flags it to you for a personal response rather than auto-posting, because those need a human judgment call on tone.
Approved responses post from your verified business account. You review the draft before it goes live; the turnaround is under 2 hours even during a lunch rush because you're approving via a simple yes/no in your phone, not writing from scratch.
Real number: Restaurant businesses that respond to more than 50% of reviews see an average Google rating 0.2-0.3 stars higher within 6 months, according to a 2023 Harvard Business School working paper on review response behavior. On a platform where 4.2 and 4.5 are psychologically very different to a diner choosing between two similar spots, 0.2 matters.
3. Schedule gap alerts
What happens now: Your FOH manager builds the schedule in 7shifts on Tuesday for the following week. By Thursday, three people have texted to swap or call out. By Saturday afternoon, you're short a server for the dinner rush. You're finding out at 4pm when someone doesn't show for their 5pm shift. You're working the floor yourself or calling in a favor from someone who resents it.
What the agent does: When a shift swap request comes into 7shifts or a confirmed schedule has a coverage gap (an open shift that's been unassigned for more than 4 hours), the agent sends broadcast availability requests to the staff pool who are qualified for that role and not already scheduled that day. "Saturday dinner, 5-11pm, server position open. Reply AVAILABLE to claim it." Whoever replies first gets the shift confirmed and the schedule updates in 7shifts automatically.
It also runs a daily 8am scan for the next 48 hours and flags any coverage gaps to the manager with current status: who's been contacted, who's confirmed, what's still open. You're not discovering coverage problems at 4pm. You're managing them at 8am.
Real number: Restaurants that go understaffed on high-volume nights see table turn time increase by 15-20 minutes (tables sit, orders take longer, guests leave without ordering dessert or a second drink). On a 60-seat restaurant doing 2 full turns on a Friday, one understaffed server position represents roughly 4-6 fewer table turns and $240-$360 less in revenue, before accounting for reduced tip income leading to higher server turnover.
4. Vendor invoice reconciliation
What happens now: Your produce distributor sends an invoice. Your fish purveyor sends an invoice. Your linen service sends one. Your beverage distributor sends three (wine, spirits, beer). Someone in your back office or you personally matches each invoice against the delivery receipt, checks quantities and pricing, and approves payment. This takes 4-6 hours per week. Errors happen. A produce invoice charges you for 20 cases of heirloom tomatoes; you received 14. You pay it anyway because it's 7pm and you're on your way out the door.
What the agent does: When an invoice arrives via email (the majority of vendor invoices come by PDF to a billing email), the agent extracts the line items, quantities, and per-unit pricing using document parsing. It compares them against your most recent delivery receipts (from Toast, from manual entry, or from your receiving log) and your contracted pricing. Any discrepancy over a threshold you set (typically $25 or 3%) gets flagged with a draft dispute email and the specific line item that doesn't match.
You don't review every invoice. You review the flagged ones.
Real number: The National Restaurant Association estimates invoice errors affect 15-20% of food and beverage invoices, with an average discrepancy of $60-$120 per error. A restaurant doing $80,000/month in food and beverage purchasing sees 3-6 errors per month. Catching half of them recovers $90-$360 per month, which doesn't sound large until you annualize it and notice it covers the cost of the automation itself several times over.
5. Post-visit review request
What happens now: A guest has a great meal. They mention it to a friend. They never leave a review because no one asked. A month later you're wondering why your food costs are right but your Google rating isn't moving.
What the agent does: When a check is closed in Toast for a table of 2 or more (single diners on quick visits are typically excluded, configurable), the agent triggers a text message to the primary reservation holder 3 hours later: "Thanks for coming in tonight. If you have a moment, we'd love a review. It helps a lot: [Google link]. Hope to see you back soon." If the guest came via OpenTable or Resy and left contact info, that data is used. If they paid by card and enabled receipt texts, the card receipt phone is used where legally permitted.
The timing is deliberate. Three hours is after the meal has settled, after they've told a friend how good it was, but before they've forgotten. Same-night while still at dinner is too pushy. Next morning is already a different day.
Real number: Proactively requested reviews convert at 15-25% of contacts, compared to 2-4% organic review rate (guests who decide on their own to write one). For a restaurant doing 100 covers per night, even a 15% conversion on review requests generates 15 new reviews per night for the first month, after which the rate normalizes as the review base grows. Restaurants in the top quartile for Google review volume in their local market see 2-3x the organic discovery traffic of those in the bottom quartile.
What it costs
A single workflow runs $2,495 to $4,995 depending on complexity and what systems it needs to touch. The review response and post-visit review request workflows are on the lower end. The invoice reconciliation, which requires document parsing and contract rate matching, is on the higher end. Fixed price, not SaaS subscriptions for something you half-configure and never fully use.
The AI Receptionist tier at $4,995 setup plus $497/month covers the full stack: no-show reduction, schedule alerts, and review management running as a continuous loop with ongoing model tuning.
Integration note: OpenTable and Resy both have webhook and API access through their partner programs. Toast POS has a robust API for transaction data and customer records. 7shifts has schedule and shift management API access. If you're on a POS system other than Toast (Square, Aloha, Lightspeed), let us know at the start. Some work without modification, some need a custom layer, and we'll scope it honestly before kickoff.
What to automate first
If empty tables are the problem, start with no-show reduction. It pays back the fastest because every recovered table is revenue that was already booked.
If reputation and local search are the constraint (you have good guests but a weak review profile), pair the post-visit review request with the review response automation. They work better together: more reviews coming in plus every one responded to compounds faster.
Don't run invoice reconciliation at the same time as a POS integration change. Two things touching your financial data layer simultaneously creates reconciliation headaches that are hard to untangle.
Free 3-minute audit
The free audit takes 8 questions about your restaurant (reservation platform, cover count, biggest operational headache) and outputs which workflow recovers the most for your specific situation. Most restaurant owners who take it are surprised to find that invoice reconciliation recovers more than they expected and that the review workflow is simpler to build than they assumed.