Skip to content
Voice AgentBible

AI answering service for restaurants: hours, location, menu and dietary questions

What an AI answering service for restaurants does with hours, directions, menu and dietary questions, when it transfers, plus KPIs and a demo script.

By · 4 min read

Last verified 01 Oct 2026v1.0Published 01 Oct 2026

Restaurants · #4 of 5Status and FAQInboundStartertarget ≤ 800 ms turn

KPIs at a glance

Key performance indicators with baseline, target and how to measure
KPITypical baselineTargetHow to measure
FAQ containmentNot applicable before deployment; tag one week of calls by intent first80-90% of hours, location, menu and policy questions answered with no transfer (rule of thumb; venues with complex menus sit lower)Calls ending without transfer or message / calls with an FAQ intent, weekly.
Answer correctnessStaff answers to hours and menu questions are rarely audited; the holiday-hours error is the classicZero wrong answers on hours, address and prices in a weekly 50-call sampleTranscript sample checked against the hours source, the POS menu and the policy sheet.
Allergen handling complianceNot applicable100% of allergen questions answered from ingredient data with the configured wording; zero guaranteesTranscript search for allergen terms, weekly review; any guarantee phrase is a defect.
Transfer accuracyNot applicableOver 95% of transfers go to the right destination per your rules (orders, bookings, manager, events), with context passedTransfers by destination against the rule for the stated intent, weekly.
Answer rate during serviceTrade press claims about 40% of restaurant calls go unanswered, unsourced upstream; log your own by half-hour99% of offered calls answered within three ringsCarrier or platform logs, answered / offered.

What it is

An answering agent for a restaurant picks up the calls that are questions, not transactions: are you open now, are you open on the holiday, where are you, is there parking, do you have a kids' menu, is the curry spicy, is anything vegan, is the bread gluten free, do you take large groups, can I bring a cake. It answers from data you already maintain, resolves dates aloud when the question is about a day, and transfers or takes a message when the question is outside its data or the caller asks for a person. When the call turns into an order or a booking it either handles that too or routes it by your rules.

These calls are short, often under a minute, and they arrive in the same hour as the orders. They are also where a prompt-only agent embarrasses you: holiday hours from last year, a price from the old menu, or a cheerful "yes, that's gluten free" about a dish that is not. The fix is structural, not verbal. Hours from the profile you already update, the menu and prices from the POS, dietary flags from the same ingredient data the kitchen uses, and policies from one sheet.

Independents call this an AI answering service or AI receptionist. Groups call it front-of-house call deflection and measure it per venue.

Who buys it

  • Independents and small groups whose staff answer the same twelve questions sixty times on a Saturday, and whose voicemail fills with "are you open?"
  • Restaurant groups with central marketing that wants one accurate answer for hours, parking and policies across venues, and a transfer tree that lands on the right person at each site.
  • Venues without an ordering or reservation integration yet, buying the questions layer first and adding orders and bookings later.

Budget owner: the owner or general manager; in groups, the marketing or operations lead, with whoever maintains hours and menu data owning the sources.

KPIs

Tag one week of calls by intent before you deploy; the share that are pure questions is the ceiling for this use case. Then track FAQ containment, answer correctness against your sources, allergen handling, transfer accuracy and answer rate during service.

The trap is correctness. Containment of 90 percent with wrong holiday hours in two percent of answers is worse than a transfer. Sample fifty transcripts a week against the sources, and treat any allergen guarantee as a defect regardless of how the call ended.

Demo script

Give the vendor your hours with one holiday exception, your POS menu with dietary flags, a policy sheet (cake, dogs, vouchers, corkage) and a transfer tree. Change one hour and one price mid-demo.

  1. Greeting. Pass: restaurant name, AI disclosure, recording notice, "how can I help?"
  2. Ambiguous date. Ask "are you open on the Monday?" when a holiday is ten days out. Pass: it names the date, reads the exception from the hours source, and offers the next open day if closed. Fail: it reads standard Monday hours.
  3. Interruption. While it recites the address, cut in with "is there parking?" Pass: it stops and answers parking from the policy sheet.
  4. Dietary question. Ask what is vegan on the menu. Pass: items read from dietary flags, with prices from the POS.
  5. Allergy question. Ask whether the bread is safe for a coeliac. Pass: ingredients and your wording, no guarantee, offer to note it on a booking or transfer. Fail: "yes, it's gluten free, no problem."
  6. Changed data. Have your administrator change a price and an opening hour in the sandbox, then ask again. Pass: the new values. Fail: the old ones.
  7. Quantity-style change. Ask "a table for four tonight... actually six, around eight". Pass: it either books with the corrected size by your rules or transfers to bookings with the corrected size and time in the note.
  8. Eight seconds of silence. Pass: one short prompt, then a graceful close with the option to call back.
  9. Out-of-scope. Ask whether you can film a music video in the dining room. Pass: "I don't have that information", then a message for the manager. Fail: an improvised yes.
  10. Transfer. Ask for the manager about an invoice. Pass: transfer to the configured destination with the question passed along, and a message taken if nobody answers.

Score each trap pass or fail. A wrong hour after the change in turn 6, or a guarantee in turn 5, is a hard stop.

Compliance notes

This is inbound and informational, and a call the customer places is outside the TCPA consent rule in the United States. Recording is the live question: roughly a dozen states require all-party consent, so announce it at the start. Several states have bot-disclosure laws, so disclose by default. On allergens, federal law defines nine major allergens, with sesame since 1 January 2023; the labelling rule does not cover restaurant food served to order, but a wrong spoken answer is still yours, so the agent reads data and never guarantees. If the agent offers to text the menu or directions, that is an outbound text and needs consent captured on the call. In the United Kingdom, food businesses must provide information on 14 allergens for non-prepacked food, and since 1 October 2021 prepacked-for-direct-sale items carry full ingredient labelling under Natasha's Law; recording is personal-data processing under UK GDPR, and any later marketing call needs specific prior consent under PECR Regulation 19. In the European Union, the AI Act's Article 50 duty to tell callers they are talking to an AI applies from 2 August 2026, and the 14-allergen duty applies. In Australia, state surveillance-devices laws differ, so announce recording; promotional callbacks sit inside the telemarketing hours of Monday to Friday 9 a.m. to 8 p.m. and Saturday 9 a.m. to 5 p.m. Informational, not legal advice; the compliance rows on this page carry the sources.

Build or buy

Buy, and buy cheaply; this is the starter use case and most ordering and reservation agents include it. The one thing to insist on is sources over prompts: hours from a profile you maintain, menu and prices from the POS, dietary flags from data, policies from one sheet you can edit. If a vendor's answer to "how do we change the holiday hours?" is "send us an email", keep looking. The acceptance test is turn 6 of the script: change a price and an hour, ask again, hear the new values.

Questions to ask vendors

  1. 01

    Where do the agent's answers about hours, menu and prices come from, and what happens when we change them?

    A good answer: Hours from a source you already maintain, the menu and prices from the POS, policies from a sheet you edit. Changes appear without a vendor ticket. Not a prompt someone pasted three months ago.

  2. 02

    What does the agent say when asked whether a dish is gluten free, vegan or safe for a nut allergy?

    A good answer: It reads the dietary flags and ingredients from your data, speaks your configured wording, and never guarantees. If the data does not cover the question it offers to transfer or take a message.

  3. 03

    How does it handle holiday hours and a question like 'are you open on the bank holiday Monday?'

    A good answer: It resolves the date aloud and reads the exception from the hours source. Shown by changing the sandbox hours and asking again.

  4. 04

    What are the transfer rules, and what does the person receiving the call hear or see?

    A good answer: Rules by intent and time of day that you set, a whisper or screen note with the caller's question, and a message option when nobody picks up.

  5. 05

    What happens when a caller asks something you have no data for, like whether the patio is dog friendly or whether you take a certain voucher?

    A good answer: A plain 'I don't have that information', then a transfer or message. It does not improvise.

  6. 06

    How does the agent behave when the caller interrupts, goes quiet, or when the menu lookup is slow?

    A good answer: Stops mid-sentence on interruption, prompts once on silence, and says it is checking rather than going silent on a slow lookup. Shown in a transcript.

Matrix rows that apply

Rows from the global compliance matrix that apply to this page. Informational only, not legal advice; dates change, confirm with counsel and the regulator.

JurisdictionConsent for automated callsAI disclosureCalling hoursRecordingVerified
United States (federal)confidence high
Required

The FCC's February 2024 declaratory ruling confirms that AI-generated or cloned voices are "artificial or prerecorded" voices under the TCPA. Outbound calls using them need prior express consent; marketing calls to mobile numbers need prior express written consent. Inbound calls initiated by the consumer are outside this consent rule.

Conditional

No federal statute yet requires an agent to announce that it is AI. TCPA rules already require prerecorded or artificial-voice calls to identify the caller at the start and give a callback number. An FCC proposal (2024) would add an explicit AI disclosure; several states have their own bot-disclosure laws. Disclose by default.

Required

Telephone solicitations only between 8 a.m. and 9 p.m. in the called party's local time (47 CFR 64.1200(c)(1)).

Conditional

Federal law is one-party consent; roughly a dozen states (including California, Florida, Washington and Pennsylvania) require all-party consent. Announce recording at the start of every call unless counsel confirms otherwise.

2026-09-30
United Kingdomconfidence medium
Required

The ICO treats conversational AI voice calls as automated calls under PECR Regulation 19, so direct marketing by automated call needs the recipient's specific prior consent. Live human marketing calls follow the softer Regulation 21 rules (screen against the TPS).

Recommended

No UK statute mandates announcing an AI caller, but PECR requires automated marketing calls to identify the sender and provide a contact address, and UK GDPR transparency duties apply.

Recommended

No statutory hours in PECR; Ofcom and industry codes expect reasonable hours and honouring "do not call again" requests.

Required

Recording is processing of personal data under UK GDPR; tell callers at the start and document the lawful basis. Financial firms have additional FCA recording duties.

2026-09-30
European Unionconfidence medium
Required

Automated calling systems without human intervention for direct marketing need prior consent under the ePrivacy Directive (Art. 13) as transposed by each member state; GDPR requires a lawful basis for the processing itself.

Required

EU AI Act Article 50 requires that people interacting with an AI system are informed they are doing so unless it is obvious. Transparency obligations apply from 2 August 2026. Proposed "Digital Omnibus" amendments may adjust timing or scope; verify before relying on this row.

Conditional

Set by member-state law and codes (for example, national telemarketing hour rules); no EU-wide statutory window.

Required

Recording needs a GDPR lawful basis and transparent notice at the start; several member states require all-party consent.

2026-09-30
Australiaconfidence medium
Required

Telemarketing calls must not be made to numbers on the Do Not Call Register without consent (Do Not Call Register Act 2006); research calls have narrower exemptions.

Conditional

The Telemarketing and Research Calls Industry Standard requires callers to identify themselves, the organisation and the purpose at the start. No general AI-caller law; broadcasting codes have begun requiring synthetic-voice disclosure in specific contexts.

Required

Telemarketing calls only Monday to Friday 9 a.m. to 8 p.m. and Saturday 9 a.m. to 5 p.m. local time; none on Sundays or national public holidays (Industry Standard 2017).

Conditional

State and territory surveillance-devices laws differ; several require all-party consent. Announce recording at the start.

2026-09-30
New Zealandconfidence low
Recommended

No statutory do-not-call register for voice calls; the Marketing Association's Do Not Call list is voluntary. The Privacy Act 2020 governs collection and use of personal information.

Not required

No AI-caller disclosure statute; Privacy Act transparency principles apply.

Recommended

Industry code expectations only.

Recommended

One-party consent for a participant; notify callers to satisfy Privacy Act collection principles.

2026-09-30

    Frequently asked

    What is the difference between an AI answering service and an AI receptionist for a restaurant?

    Vendors use both names for both things. On this site, an answering service answers questions and takes messages; a receptionist also writes orders and bookings into your systems. Most restaurants want the second; buy the first only if you have no POS or reservation integration available.

    Can the agent answer allergen and dietary questions?

    It can read ingredient and dietary data you provide and speak your wording. It should never promise a dish is safe, because cross-contact happens in kitchens. In the UK and EU you have a legal duty to provide information on 14 allergens; in the US nine are defined for labelling. Test it with your own recorded call.

    How often do the answers go stale?

    As often as your sources do. If hours live in a profile you already update and the menu comes from the POS, the agent stays current. If they live in a prompt, every holiday is a risk.