An AI host for restaurants and cafes
that books tables and never forgets an allergy question

Where a restaurant loses guests without noticing

A restaurant’s phone rings hardest exactly when nobody is free to answer it: during service, with the floor full and the kitchen behind. Missed-call rates on restaurant phone lines commonly run 20 to 25 percent during peak hours. A caller who cannot get through rarely calls back. They book the next place on the list, or just walk in somewhere else. Outside opening hours the gap is total, even though plenty of people decide where to eat tonight while scrolling on their phone late at night.

The questions that eat up a host’s attention are repetitive and low-risk to automate: “is there a gluten-free option”, “are you open on the holiday”, “can we get a table for six at eight”, “where is my delivery order”. None of these need a manager’s judgment, yet they interrupt service dozens of times a shift and go completely unanswered after close. Restaurants that already run a WhatsApp number informally know the volume is there. What is missing is a system that answers the same way every time, at any hour, without someone glued to a phone between tables.

There is also a cost to getting the answer slightly wrong under pressure. A rushed host guessing at an allergen, or confirming a table that was actually double-booked twenty minutes earlier, can turn a routine question into a bad review. Worse, it can become a real safety issue. Staff usually know the menu well enough. The problem is that the busiest moments are exactly when there is no time to check carefully. A phone call competes directly with the guests already standing at the host stand.

What the agent handles at the host stand

A booking agent that reads your real availability. It checks actual open tables in your reservation system or calendar. It books, reschedules and cancels, and never promises a table that does not exist. Menu, allergen, pricing and hours questions get answered from a knowledge base built from your actual menu and policies, not a generic script that gets ingredients wrong.

Order status without a phone call. For restaurants running delivery or pickup, the agent answers order-status questions from your order system in real time. That beats a guest calling the one phone line that is also taking reservations.

Waitlist and no-show recovery. A table can free up from a cancellation at any moment. The agent offers it to guests on a waitlist, often before the host even notices the gap. Automatic reminders before a reservation also cut the no-shows that leave a table empty on a busy night.

Feedback triage. A guest who says something positive gets asked for a review on the platform that matters to you. A guest with a complaint gets routed straight to a manager, with the conversation attached. That way a bad night does not become a bad public review before anyone on your team even hears about it.

Integrations that fit a restaurant. OpenTable-style booking platforms, Google Calendar or your POS handle scheduling and orders. WhatsApp, Telegram, Instagram Direct or a website widget carry the conversation. Your existing CRM or spreadsheet works too, if that is what you already run on.

A human for anything sensitive. Severe allergy questions, complaints and large private events go straight to your staff. So does anything that needs real judgment. The full conversation goes with the handoff, rather than being answered by the agent.

A knowledge base that stays current. Menus change, dishes sell out, and a seasonal item appears for six weeks and disappears. The agent reads from a knowledge base your team can edit without a developer. A weekend 86 on the salmon shows up in the agent’s answers the same day, not weeks later when someone remembers to update a script.

Weeks 1 to 4: how we build it

1. Week 1: audit and playbook. We sit with your front-of-house team and collect your menu, allergen list and the questions they field most. That becomes the agent’s knowledge base and escalation rules.
2. Week 1-2: reservation, POS and channel integration. We connect to your booking platform, calendar or POS and set up the messenger channels you choose.
3. Week 2-3: build and internal testing. The agent runs against real booking and order scenarios and a set of test conversations before any guest sees it.
4. Week 3: soft launch. The agent goes live for a subset of conversations, or during off-hours only, while your team reviews what it says.
5. Week 3-4: full launch and tuning. Full rollout, with two weeks of close review and adjustment included in the package.

What it costs

Prices follow the AI agent service packages. The exact figure depends on your reservation system and how many channels you need.

What restaurants typically see

Restaurants running a booking and FAQ agent typically see 50 to 70 percent of routine questions answered without staff involvement. That is an industry benchmark range, not a guarantee for any specific venue. No-show rates commonly drop when automatic reminders and one-tap rescheduling are in place. A guest who intends to cancel usually does it early enough for the table to be refilled from a waitlist. Booking and order-status coverage becomes 24/7, instead of service hours only. That matters most for late-evening decisions, which otherwise go to whichever competitor answers first.

Response speed also tends to matter more than most owners expect. The window for keeping an inquiry warm is measured in minutes, not hours. A guest who gets an instant, accurate answer about tonight’s availability is far more likely to book on the spot than one left waiting for a callback. None of this is a promise of a specific lift for your restaurant. It is the pattern we see across the category. The agent is built to be measured honestly against your own numbers, once it is live. Our own numbers are in the case studies linked below.

Why work with us

We build these agents the way we build them for our own businesses: one job per agent, a closed list of tools it can use. A human approves anything risky, like a complaint or a severe allergy question. Pricing is fixed before work starts, not billed by the hour, so there are no surprises on the invoice. You get a working demo every week, not a single reveal at the end. After launch we keep tuning the agent with you for two weeks as part of the package, instead of handing over a system and disappearing.

We also do not oversell what an agent can do for a small, single-location cafe. If your volume of repeat questions is genuinely low, we will say so during the audit. We would rather suggest a simpler setup than sell a full agent package nobody needs.

An AI host is one half of the picture for most restaurants. Many pair it with a Telegram bot for reservations or delivery. Or look at the wider AI agents and automation service for what else can run the same way. The same playbook approach built a multi-channel sales agent across seven platforms. The same human hand-off discipline runs a visa center’s support bots.

Tell us about your reservation system and channels and we will send back a fixed price and a two-to-four-week-to-launch plan: get in touch.