TALON — AI receptionist

It answers. It books.You draw the line.

TALON picks up your business number, plays through the booking, and drafts replies on email, messaging and chat. You sketch the rules — alone, ask you, or never — under a monthly AI spend cap, with a full record of every move.

Talk to the teamSee the map
Generative constellation of front-desk moves around the TALON mark.
The TALON console dashboard in the dark theme: a live-call banner with a running timer, cards for pending approvals, open conversations, upcoming bookings and active leads, and the month’s numbers — calls answered, answer rate, bookings made by the AI, estimated revenue captured.
A front desk that generates next steps · The console, during a call — a screenshot, not a mockup
How it works

A small diagram of a big front desk.

One request comes in. TALON routes it, checks the calendar, drafts the reply, and stops when your rules say so. Play with the map — every step links to a chapter.

Schematic map: a caller reaches TALON, which checks the calendar and your rules, then either acts or asks you.
The route — schematic, not a mockup
  1. It hears the ask Phone, mail, messaging or the chat on your site — same intent, different door.
  2. It walks the call Disclosure, a real slot, a confirmation, and the moment it hands the decision back.
  3. You keep the verdict Autonomy matrix, spend cap, one approvals queue. Nothing leaves without your frame.
  4. It wears your brand Booking page and chat widget in your colour, your name, your tone.

Measured, not claimed

Four things you can check.

The brain
Lodestar

Every answer is produced on Altovar’s own AI gateway, in the EU. In production the direct path to model vendors is closed: if the gateway does not answer, TALON does not go shopping elsewhere.

Your rules
Autonomous · approval · forbidden

For each action type and each level of stakes you pick one of three verdicts: it acts alone, it waits for you, or it never does it. Email and messaging replies pass through that matrix before anything leaves.

Your spend
A monthly cap, in euro

Past the warning threshold, autonomous sends start queuing for your approval; past the cap, they stop. The spend is visible per agent.

Double bookings
Checked in the transaction

Overlaps and covers are re-checked inside the same transaction that writes the booking. Two calls for the last table cannot both win it.

Four ways a call can go.

Pick a situation. The words are a sample; what TALON does with them is what the product does today, verified in the code that runs it.

Live
  1. The caller

    Hi, do you have a table for four on Saturday evening?

  2. TALON

    Good evening. I am an assistant, not a person. Saturday at 8:30 pm I have a table for four: shall I confirm it?

  3. The caller

    Yes, under Rossi.

  4. TALON

    Done: Saturday, 8:30 pm, four people. The confirmation is on its way to your email.

What it did

It answers on your number
A real number, activated from the console. TALON answers, says it is an assistant, and speaks the caller’s language among the five it knows.
It books while talking
It offers only slots that are free — covers and overlaps counted — and writes the booking in the same transaction that checks it.
It leaves a full record
Summary, transcript, the data it collected, the outcome and how the call scored against your criteria — and the recording, playable in the console.
Live
  1. The caller

    It’s Bianchi, I booked with you last week.

  2. TALON

    Welcome back. I see Tuesday’s booking: the same time, on Thursday?

  3. The caller

    Thursday is fine.

  4. TALON

    Thursday at 7 pm is free: confirmed.

What it did

It recognises who calls back
A caller who already booked or already called is known: name, last booking and last summary are at hand during the call.
It books while talking
It offers only slots that are free — covers and overlaps counted — and writes the booking in the same transaction that checks it.
It leaves a full record
Summary, transcript, the data it collected, the outcome and how the call scored against your criteria — and the recording, playable in the console.
Live
  1. The caller

    I wanted to know if you deliver at home.

  2. TALON

    I don’t have that information. I’ll take a message and someone will call you back.

  3. The caller

    Fine, by tomorrow please.

Urgency: mediumWaiting for your acknowledgement

What it did

It takes the message when it should
When there is nothing to answer with, it takes a message: transcribed, flagged low, medium or high urgency, waiting for your acknowledgement.
It leaves a full record
Summary, transcript, the data it collected, the outcome and how the call scored against your criteria — and the recording, playable in the console.
Missed call

  1. 12:41 — the call could not be answered.

  2. SMS sent from your number

    We couldn’t pick up. You can book from here: your booking link

What it did

It texts back a missed call
A call it cannot take gets an SMS with your booking link. Needs SMS switched on for the number.
Schematic of a live call: TALON on the line with four outcome cards beside the handset.
On the phone — schematic
The call

What happens when it rings.

A booking taken over the phone is a chain of small decisions. Here is the order they happen in — and the point where TALON stops deciding.

  1. It picks up, and says what it is

    Your number rings and TALON answers, introducing itself as an assistant. That disclosure is built into the product, not a setting you remember to switch on.

  2. It checks before it promises

    Availability comes from your calendar, covers and overlaps included. What it offers is what is actually free — and when it writes the booking, it checks once more inside the same transaction.

  3. It confirms, and keeps confirming

    The guest gets an email with a manage link at once, a reminder the day before, a text an hour before where SMS is on. Moving the booking is self-service; not showing up is recorded on its own.

  4. When it shouldn’t decide, it doesn’t

    Nothing in your data to answer with? It takes a message — transcribed, with an urgency level — instead of improvising. On the written channels every reply queues for your approval; on the phone it acts in real time and leaves you the full record.

The autonomy is yours to set: which actions are forbidden, which need your approval, which run on their own. Whatever you choose, every action is logged with its cost.

Four-step call diagram ending in a hand-back stamp for your decision.
The call — four moves

One request, three written channels.

The same message can arrive by email, on messaging or from the chat on your site. Switch channel, then decide how much it may do on its own.

Customer

Good morning, is there room for four on Saturday evening? Around half past eight.

TALON’s draft

Good morning! Saturday evening there is a table for four at these times. Which one shall I hold for you?

8:30 pm 9:00 pm Real slots, read from the calendar when the draft is written.
It sends on its own
Autonomy on this channel
It sends on its own

Sent. The conversation goes on without you and stays on the record.

Connect the mailbox you already use — OAuth or IMAP. TALON skips spam and auto-replies, spots booking intent and drafts answers with real slots already in them.

Customer

Good morning, is there room for four on Saturday evening? Around half past eight.

TALON’s draft

Good morning! Saturday evening there is a table for four at these times. Which one shall I hold for you?

8:30 pm 9:00 pm Real slots, read from the calendar when the draft is written.
It sends on its own
Autonomy on this channel
It sends on its own

Sent. The conversation goes on without you and stays on the record.

Scan a QR code from the business phone and the channel is on — no separate business account. Drafts there follow the same rules and the same approvals.

Customer

Good morning, is there room for four on Saturday evening? Around half past eight.

TALON’s draft

Good morning! Saturday evening there is a table for four at these times. Which one shall I hold for you?

8:30 pm 9:00 pm Real slots, read from the calendar when the draft is written.
Chat on your site
Autonomy on this channel

In the chat it answers straight away, from your knowledge base, and says it is an assistant.

One script tag, your colour, your assistant’s name. It answers from your knowledge base and says it is an assistant, on every visit.

On every written channel

It answers from your own articles
Every channel draws on the knowledge base you write. What it quotes is the summary you give each article — your words, not its guesses.
Tickets, when a reply is not enough
A helpdesk underneath: kanban, canned responses, SLA timers with escalation, satisfaction surveys and ticket analytics.
Manners built in
One-click unsubscribe on commercial sends, a suppression list it never writes to, and rate limits on every public surface.
Three written channels side by side: email, messaging and site chat, each with a draft frame.
Written paths — schematic

A booking, from the first minute to the day after.

Move along the timeline. Every step is something TALON does on its own once the booking is written.

Before the first booking

Services, your way
Duration, buffers, booking window, auto-confirm, price, capacity per slot, and custom questions the guest answers while booking.
A public page under your name
One page per service, with your logo and your colour, and an anti-bot check on every booking.

Written

Written in the same transaction that checked the slot: capacity per slot and overlaps counted, whether it came from the phone, an email or the public page.

Confirmed

An email the moment the booking is written. From that link the guest moves or cancels on their own, without calling back.

The day before

A reminder email goes out on its own.

One hour before

A text message, where SMS is switched on for the number.

Fifteen minutes after

If nobody showed up, the booking is marked a no-show on its own, fifteen minutes after the slot ends. No list to go through by hand.

The follow-up

A no-show or a completed visit can start a sequence you designed — email, SMS, waits and conditions.

Booking timeline with five beats from write to follow-up.
Booking — from minute zero

More than one agent, and the numbers to read.

Each inbound line goes to the agent you choose, with its own persona, hours, autonomy, knowledge scope and budget. Pick one to see where it lands.

Routed to

Front desk

Hours
24/7
Autonomy
Answers and books on its own
Knowledge scope
Opening hours, services, bookings
Monthly cap
€ 40 Spend tracked per agent
Routed to

Sales

Hours
Mon–Fri, 9 to 18
Autonomy
Drafts wait for approval
Knowledge scope
Price list and quotes
Monthly cap
€ 120 Spend tracked per agent
Routed to

Sales

Hours
Mon–Fri, 9 to 18
Autonomy
Drafts wait for approval
Knowledge scope
Price list and quotes
Monthly cap
€ 120 Spend tracked per agent
Routed to

Front desk

Hours
24/7
Autonomy
Answers and books on its own
Knowledge scope
Opening hours, services, bookings
Monthly cap
€ 40 Spend tracked per agent

The autonomy matrix, the approvals queue and the two budget thresholds are shown further down, in Control. Control

And to grow on

Leads with a score
Every inbound email can become a lead, scored 0–100 on source, engagement, fit, recency and tone, on a kanban from new to won.
Sequences
Email, SMS, waits and conditions, started by a new lead, a no-show or a completed visit. They go out in working hours; outside them they wait.
An API with a spec
A public REST API with an OpenAPI document and self-service keys — tickets, bookings, leads, articles, conversations. Webhooks and team-chat notifications for the events you pick.
Numbers you can read
Calls answered, answer rate, bookings made by the AI and the revenue they represent, month to date. Ticket analytics with SLA compliance and agent performance.

Verified path by path in the code that runs it, on 1 September 2026. Nothing on this page is planned or coming soon.

The telephone, SMS reminders, the lead pipeline and sequences depend on the plan. Everything else is in every plan.

Growth router: four inbound lines converging on one agent, then out.
Growth — who answers what

Every call, on the record.

TALON acts while you cannot answer. The console shows you all of it — while it happens, and after.

Live, then on the record
A call in progress is announced on every page, timer running. When it ends, the summary, the transcript and the collected data are attached to the call itself.
Scored on your criteria
Booking made, message taken, question answered: each call is evaluated against explicit criteria, and the verdict sits next to the transcript, not in a report somewhere else.
A number becomes a person
Names resolve from your bookings and leads — on the calls list, on the voicemail list and on the call itself.
A call record in the TALON console: the collected data (party size and customer name), the evaluation badge “Appointment booked”, and the transcript of the exchange between the assistant and the caller. A call record in the TALON console: the collected data (party size and customer name), the evaluation badge “Appointment booked”, and the transcript of the exchange between the assistant and the caller.
A call record — collected data, evaluation, transcript
Control

You decide what it may do.

An assistant that can send emails and book tables needs rules, not trust. TALON’s are explicit, per action, and enforced on the server.

Autonomous, approval, forbidden
Per action type and level of stakes. A low-stakes email reply can go out alone while a high-stakes one waits; a refund can be forbidden outright.
One queue for what waits
Each draft comes with the AI’s intent, confidence and reason. Approve and send, edit first, or reject — from one screen.
A budget with two thresholds
Set the monthly cap. At the warning threshold autonomous sends start queuing; at the cap they stop. Spend is tracked per agent.
The autonomy matrix in TALON settings: rows for replying to emails, sending new emails, replying on mobile chat, creating bookings and cancelling bookings, each with a verdict — autonomous, needs approval or forbidden — per level of stakes. The autonomy matrix in TALON settings: rows for replying to emails, sending new emails, replying on mobile chat, creating bookings and cancelling bookings, each with a verdict — autonomous, needs approval or forbidden — per level of stakes.
Settings → Autonomy — the matrix
The approvals queue in the TALON console: two email drafts waiting, and the selected one open with the recipient, the AI’s intent and confidence, the subject and the drafted message. The approvals queue in the TALON console: two email drafts waiting, and the selected one open with the recipient, the AI’s intent and confidence, the subject and the drafted message.
The approvals queue, with a draft open
Three autonomy columns: alone, ask you, never — stacked action rows beneath each verdict.
Control — your verdicts

Your name on the front of it.

Guests see your business, not ours. TALON’s public surfaces carry your colour and your logo — and the text on top of them is computed for contrast against WCAG, not eyeballed.

The booking page
A public page per service with your logo, and your brand colour on the stepper, the buttons and the focus ring. Guests book, move and cancel there on their own.
The chat widget
One script tag, your colour, your welcome line, your assistant’s name. The text colour on your brand colour is calculated — readable on light and dark brands alike.
Your mailbox, your voice
Confirmations and replies leave from the mailbox you connect — OAuth or IMAP — in the informal voice of the five languages, with a one-click unsubscribe on commercial sends.
The public booking page of a demo workspace: the workspace logo, the service header, a three-step wizard and a week of available time slots, in the workspace’s brand colour. The public booking page of a demo workspace: the workspace logo, the service header, a three-step wizard and a week of available time slots, in the workspace’s brand colour. The TALON chat widget, open, in the demo workspace’s brand colour: a header named Assistente, the mandatory AI disclosure, a welcome message and the message box. The TALON chat widget, open, in the demo workspace’s brand colour: a header named Assistente, the mandatory AI disclosure, a welcome message and the message box.
The public booking page of a demo workspace · The chat widget, in the workspace’s colour
Branded booking page with slot cards and a chat widget in the corner.
Your brand on the public surface

What this does not do yet

Straight about the gaps.

The rest of the page is only as believable as this list.

Hand-off to a human

There is no way to set a target number and transfer the call. TALON stays on the line it answered.

Call sentiment

The product does not score how a caller feels. It records what it did, not a mood.

A mobile app

There is no TALON app on a phone. The console is the web product you already have.

Outbound dialling

TALON answers the number you give it. It does not place outbound sales calls.

TALON is one of ours.

It shares its identity layer, its AI gateway and its European footprint with the rest of the Altovar catalogue. One supplier, one place to ask, one jurisdiction.

Talk to the teamThe Altovar catalogue