Restaurant AI Phone Answering

The restaurant AI phone answering template is a pre-built Sawy agent that handles your restaurant's phone calls during the dinner rush and every other hour of the day. It takes reservations, answers menu and allergen questions, manages waitlists, handles takeout orders, and provides directions and hours — all without pulling your host or manager off the floor.

This page covers deployment: what the template ships with, the sample dialogues it handles during dinner rush, the integrations and write actions it performs in each reservation platform, the tuning tips that matter most for restaurants, and where this template is the wrong fit. For the broader market case and ROI math on AI in restaurants, see the AI phone agent for restaurants page.

Where this template fits

| Business shape | Fit | Notes | |---|---|---| | Solo operator, predictable hours | Strong fit | Template handles routine; you take exceptions | | Small team with a front desk | Strong fit | Template handles overflow + after-hours | | Multi-location with central reception | Good fit | Configure separate rules per location | | Enterprise with complex routing | Partial | Use as a base; expect heavier customization | | Highly regulated (healthcare/legal first-touch) | Limited | Use only with explicit human-handoff tier |

What This Template Ships With

This is what the Restaurant Host template adds on top of a generic Sawy install — restaurant-specific intake patterns, hospitality dialogue tuning, reservation-platform write actions, and waitlist logic that Customer Support and Appointment Booker templates don't carry.

Sample Conversation

AI: "Thank you for calling Trattoria Bella. How can I help you tonight?"

Caller: "Hi, I'd like to make a reservation for Saturday."

AI: "I'd love to help. How many guests?"

Caller: "Four people, around 7:30."

AI: "Let me check. I have 7:15 or 7:45 available for a party of four on Saturday. Which would you prefer?"

Caller: "7:45 works."

AI: "Great. Can I get a name for the reservation?"

Caller: "Anderson."

AI: "You're booked for four at 7:45 on Saturday under Anderson. Any dietary restrictions or special occasions I should note?"

Caller: "One person is gluten-free."

AI: "Noted — we have several gluten-free options and our chef can accommodate that easily. I'll text you a confirmation. Is there anything else?"

The reservation is confirmed in under a minute, and the kitchen is already flagged about the dietary need.

Integrations Included — Specific WRITE Actions

What the agent actually does in each connected system. These write actions are scoped to the restaurant host workflow — the agent is not a generic CRM client:

• OpenTable: creates the reservation with party size, time, and guest name; pulls the guest's diner profile if the phone number matches; writes allergens, seating preference, and special-occasion flag to the reservation notes; marks the appointment with the hold-window rule so the host stand sees it.
• Resy: creates the reservation with the same fields as OpenTable; for walk-in waitlist additions, writes the bar-seat alternative and the requested time into the walk-in queue with the caller's phone for text notification.
• Yelp Reservations: creates the booking with party + time + name; pushes allergen flags to the reservation notes (Yelp's schema supports fewer custom fields, so the template surfaces this as a plain-text tag).
• Toast POS: for takeout, creates the order ticket with menu items + modifiers; attaches the estimated pickup time; flags allergens at the ticket header so the line cook sees them before plating; writes the caller phone to the customer record.
• Square for Restaurants: same as Toast for order creation; payment can be configured for either pay-at-pickup or upfront capture via Square's secure handoff (the agent does not handle card numbers directly).
• Google Calendar: fallback reservation tracking writes the booking as a calendar event with party size in the title, name and contact in the description, and allergen notes in the location field for venues not on a dedicated reservations platform.
• SMS: confirmation message fires with reservation details, hold-window reminder, and directions link. For waitlist entries, a second message fires when the table is ready, with a 10-minute claim window before the slot rotates to the next caller in queue.

Setup in 3 Steps

1. Select the template — Choose Templates → Restaurant Host from your Sawy dashboard. This loads hospitality greetings, reservation flows, and menu Q&A capabilities.
2. Customize your restaurant — Add your restaurant name, hours, reservation platform credentials, menu highlights, allergen info, and special policies (dress code, corkage, large party minimums).
3. Go live — Forward your restaurant's phone line and make a test call. Book a reservation, ask about gluten-free options, and verify everything syncs.

Who It's For

• Fine dining restaurants where every missed reservation impacts revenue
• Casual dining chains handling high call volume across locations
• Fast-casual spots fielding takeout and delivery calls during rush
• Catering-focused restaurants managing event inquiries and large party bookings
• Food trucks and pop-ups that need a phone presence without a host stand

Can I configure the large-party threshold for manager routing?

Yes — the default is parties of 8+, but the threshold is a single field in the rule editor. You can also layer a kitchen-complexity rule (6-tops on Saturday night routed to the manager even though they're under the size threshold) using the conditional logic builder.

How do I configure the hold-window policy the agent verbalizes?

The hold window is a per-restaurant setting (default 15 minutes). Whatever value you set, the agent verbalizes it in every confirmation along with the consequence ("after that the table goes back to the waitlist"). You can also configure a longer hold for VIP guests or deposit-bearing large parties.

Can I disable takeout orders during peak dinner hours, or hand them to a different template?

Yes — the template supports time-windowed feature gates. A common config: "takeout 11–5 and 9–10, reservation-only 5–9." If your venue is takeout-heavy overall, the Order Taker template is built specifically for menu walk-throughs and may map better than the Restaurant Host template.

How do allergens propagate from the reservation to the kitchen?

The allergen flag is one field written at booking. It propagates as: a flagged note on the reservation, a kitchen-ticket header tag when the party arrives and the host opens their check, and a host-stand alert at seating. In Toast and Square, the field maps to the ticket-header tag; in OpenTable and Resy, it writes to reservation comments with an [ALLERGY] prefix so it stays visually distinct on the floor view.

A Second Sample Call — Saturday Night Walk-In Inquiry

A different call shape this template handles during peak service: a caller asking about walk-in availability when the floor is at 95% capacity, with a 45-minute waitlist for tables.

AI: "Trattoria Bella, thanks for calling."

Caller: "Hi — can you fit two of us in tonight? Like in the next hour?"

AI: "Let me check tonight's floor. For two people right now, our waitlist is running about 45 to 60 minutes for a table. The bar has two seats open immediately if that works — full dinner menu, same kitchen, just a quicker option."

Caller: "Bar seats are fine. Can you hold them?"

AI: "Bar seats aren't reservable but I can put your name in for the next walk-in claim if you can be here in 15 minutes. If you'd rather book a table, the soonest confirmed seating for two is 9:45 tonight, or I have 7:30 open for tomorrow."

Caller: "Let's do the 9:45 confirmed. Better than racing across town."

AI: "Confirmed for two at 9:45 tonight. Name on the reservation?"

Caller: "Walker."

AI: "Booked, Walker. We hold reservations 15 minutes — past that, the table goes back into the waitlist, so if you're running late, call us. One last thing — anything to flag? Allergies, celebrating something, seating preference?"

Caller: "Nope, just hungry."

AI: "Then we'll see you at 9:45. Confirmation text is on the way."

The AI didn't oversell the waitlist (would have created a no-show), gave the caller three real options ranked by speed, set the 15-minute hold expectation explicitly so the host isn't holding a phantom table at 10 p.m., and skipped the SMS upsell prompt because the floor is too busy to need it. The reservation lands in Resy with the correct hold-window flag.

Restaurant-Specific Tuning Tips

Three tuning choices that matter for restaurants more than for other verticals:

1. Allergens belong on the kitchen ticket, not just in the reservation notes. Configure the agent to capture allergies during booking and to propagate them as a flagged note to the POS, the host stand, and the kitchen — not just into the reservation comments where the line cook will never see them. A guest with a shellfish allergy who told the AI at booking and still gets a contaminated dish has a worse experience than if they'd never said anything.
2. Hold-window policy needs to be spoken, not just enforced. A 15-minute hold is standard, but if the agent never says "we hold for 15 minutes" the no-show rate stays the same and the host eats the lost cover. Configure the agent to verbalize the hold window in every confirmation, with the consequence ("after that the table goes back to the waitlist"). The audible policy halves the no-show rate without changing any other variable.
3. Large-party threshold needs to be honest about kitchen capacity, not just table size. A six-top is bookable on most nights; a six-top with a tasting menu request at 7:30 p.m. on a Saturday isn't, because the kitchen can't pace it without dropping ticket times across the floor. Configure large-party thresholds by both party size and kitchen complexity — Saturday-night six-tops route to the manager for confirmation rather than booking automatically.

When Not to Use This Template

This template is wrong for restaurants whose phone is mostly used for takeout and delivery orders rather than reservations — use the Order Taker template instead, which is built around menu walk-throughs and order capture. It's also wrong for very high-end restaurants where the reservations line is part of the brand experience and guests expect a maître d' by name. Fast-casual concepts and food trucks generally don't need this template either; a simple Customer Support template handles their inquiry mix without the reservation infrastructure.