Booking & Phone Agent | Real Minds AI
Retail & Hospitality /Customer-service agent live field guide · 9 min

Booking & Phone Agent

Answers calls 24/7, takes and confirms bookings from your own availability and menu information, sends reminders to cut no-shows, and routes anything complex or sensitive to a person with the call context attached. It works from your approved information and escalates rather than guessing.

theater/demos/retail-hospo_booking-phone-agent.html · sandbox · read-only
Open
FIG. 1

The live demo, running on fabricated data. Open it to step through the full flow — every output is shown for a person to approve before anything happens.

How it would work

Takes the inbound call, SMS or voicemail, transcribes and classifies it against your booking rules and menu, drafts a grounded reply, and surfaces every booking — escalating anything sensitive — for a person to approve before it is sent or confirmed.

Input 01
The inbound message

A missed call, voicemail, SMS or web enquiry — captured and transcribed — carrying the booking ask, the party size, the date and time, and any dietary note the caller leaves.

Agent 02
Classifies, routes, drafts

Reads the transcript, classifies intent and urgency, checks availability and your menu and FAQ knowledge base, and drafts a reply — flagging a severe allergy or any request it cannot confirm.

Output 03
A draft, queued for sign-off

A drafted reply and a proposed booking, with the rule and source it used shown, queued for a host or manager to approve, edit or escalate — a severe-allergy booking is held for the chef, never auto-confirmed.

Where it works well

It catches every call during service and never lets a routine enquiry sit in voicemail.

  • Best for a venue carrying real overflow — a busy single site or a small group fielding bookings, modifications and hours questions all night.
  • Routine asks — hours, location, a standard table for four — get a grounded answer or a confirmed booking from your own availability, not a model's guess.
  • At hundreds of overflow calls a month, the recaptured hours go back to the guest at the host stand and to chasing the no-shows worth chasing.

The invisible cost is the call that rings out during a dinner rush — the host is with a guest, the phone goes to voicemail, the table sits empty, and the no-show that a reminder would have caught is never chased.

Where it works badly

It is confidently wrong when your menu, hours or availability behind it are stale.

  • Weak on the off-script call — a large function, a supplier, an angry guest, a media enquiry. It should route to a person, not improvise.
  • A dietary or allergen question is exactly where a confident wrong answer is most dangerous, so it holds and escalates rather than reassuring the caller itself.
  • If most of your calls are bespoke negotiations rather than standard bookings, it deflects fewer than you'd hope and the human still takes the call.
The honest test

If you cannot say, right now, that your menu, your trading hours and your live availability are all current — this tool tells your caller the wrong thing faster, not better.

Point it at a menu that still lists last season's dishes, trading hours that changed for a public holiday, or a booking system it can't see live, and it answers a caller cleanly and incorrectly. That is the trap.

What it doesn't do — and shouldn't

It drafts and books the routine. A person approves anything sensitive. That boundary is deliberate.

WHAT IT DOES
Surfaces the intent, party size, date and time it captured from the call
Shows the menu line, hours or availability slot it answered from
Flags a severe allergy, a complaint or any request it could not confirm
WHAT IT WON’T
Confirm a booking that carries a severe-allergy or dietary-safety note on its own
Promise the kitchen can safely cater for an allergen
Handle a complaint, a refund or a dispute without a person

Catering for a severe allergy is a food-safety obligation under the Food Standards Code (Standard 1.2.3) and the state or territory Food Act your venue trades under, and a wrong reassurance can put a guest in hospital. A booking is also a representation under the Australian Consumer Law. The accountable person stays on the decision because the consequence — to the guest and to the venue — lands on them, not the tool.

What your data has to look like

A live booking system, a current menu and FAQ, and trading hours the agent can actually read.

44%
Typical readiness
across orgs we see, before the first job
Live availability from your booking system
Needs shaping
A current, human-edited menu and allergen knowledge base
Needs shaping
Trading hours, including public-holiday and seasonal changes
Usual weak point
Clear routing and escalation rules
Usual weak point
A reminder and contact channel (SMS / phone)
Usually ready
The real first job

The weak point is almost always that availability and the menu aren't live — the booking system isn't connected, or the allergen notes live in a printout the kitchen keeps in its head. Getting that data captured, connected and kept current is usually the real first job — larger and more valuable than the voice layer on top. Once the inputs are live and correct, every call after that is answered right by default.

Right fit if…
You carry real overflow — calls ring out during service and bookings go to voicemail
Mostly standard bookings, modifications, hours and location questions
Your booking system exposes live availability the agent can read
You can point to a current menu and allergen list, kept up to date
Walk away if…
Most calls are bespoke negotiations — large functions, suppliers, disputes
Your menu and hours live in a PDF or a printout nobody updates
Your booking system can't be read live, so availability is guesswork
You want it to handle allergies, complaints or refunds without a person
Open questions

The worried-buyer questions, answered straight

No — that exact case is held, not answered. When the agent detects a severe allergy or anaphylaxis note, it drafts a holding reply, does not confirm the booking, and escalates to the manager and head chef for a call-back. Catering for an allergen is a food-safety obligation under the Food Standards Code (Standard 1.2.3) and your state or territory Food Act, so a person confirms the kitchen can keep the guest safe before anything is committed. The tool surfaces the request; a person stands behind it.
It answers from whatever menu, allergen list and trading hours you give it — so if those are stale, it answers a caller cleanly and incorrectly, and it can’t tell an out-of-date dish from a current one. For a retail-hospitality venue, getting the menu, allergen notes and hours into a current, human-edited knowledge base — and the booking system connected live — is usually the first piece of work, and the piece that pays off on every call after.
No. It takes the overflow calls that currently ring out during service — the routine bookings, modifications and hours questions — so your host stays with the guest in front of them instead of choosing between the guest and the phone. The judgement calls — a function, an allergy, an upset guest — still route to a person with the call context attached. The capacity it frees goes back into hosting, not headcount.
Current to today’s service. Availability has to be live from your booking system or the agent double-books; trading hours change for public holidays and seasons; and a menu or allergen note that’s a week old can be wrong. The honest test: can you say, right now, that your menu, your hours and your live availability are all current? If not, the agent answers confidently from stale data.
A caller’s name, number, booking and any dietary note are personal information under the Privacy Act, and call recordings carry their own state and territory consent rules. Any deployment runs against your own systems and booking data, not a shared pool — we scope where recordings and transcripts sit, retention, and who can see them as part of the build. The demo here runs entirely on fabricated data; Megan Doyle is not a real caller.
It can, which is why it’s scoped to deflect the routine and escalate everything else rather than improvise. It answers hours, location and standard bookings from your own approved information, and routes a complaint, a function or an allergy to a person — it never argues with an upset guest or makes a promise the venue can’t keep. You also can’t pad your own reviews to cover a bad call: manipulating online reviews breaches the Australian Consumer Law, which the ACCC enforces.
What it takes to build
4–6 weeks · 4 phases
Reused from template~65%
Bespoke to this skin~35%
stack · Voice AI · Claude · booking API · RAG menu/FAQ · review UI
What it would cost

Fixed scope, fixed price, fixed dates.

01
Bite-sized first piece
One contained change, low risk
02
Pilot build
Most builds land here
03
Embedded support
Scale on proof

Considering this for your venue?

The honest place to start is a bite-sized first piece — one contained change, low risk. Tell us where the calls are falling through; we'll play it back, scope it, and show you what's possible.

More in Retail & Hospitality
Timesheet-to-Payroll Reconciler
View →
Shipment-Status Responder
View →
Demand-Matched Roster Builder
View →
Spec & Allergen Concierge
View →
How We Work Proof Talk to us
How We Work Proof Talk to us
Ask us anything