
AI agent for hospitality and bookings: a restaurant, tourists and after hours
Simulated scenario of a small restaurant: bookings by phone and WhatsApp, tourists in several languages, the maths in the open, setup and real limits.

What a small private clinic's day could look like with an AI agent for appointments and reminders: setup, an ROI calculation with stated assumptions, honest limits.
Simulated scenario: this is not a real customer or a success story. The clinic in this article does not exist, has no name, and its numbers are assumptions we state with the calculation in plain view. What is real is what AgenteUno does and does not do, which we checked in the code and in the product documentation. If you want to know why the product exists, we explain it in why AgenteUno was born.
Picture a small private clinic: two practitioners, one person on reception covering mornings and afternoons Monday to Friday, and around 200 appointments a month. It is a deliberately ordinary scenario.
Its day looks like many others:

An AI agent does not fix the clinic, but it can cover three concrete administrative tasks: answer when reception cannot, book the appointment and remind about it. Nothing more, and that is exactly the boundary we defend in this article.
This conversation is written as an example, it is not from any real patient. It shows the behaviour you would configure, not a guarantee that the model will always answer in these words.
Patient (21:40, on WhatsApp): Hi, good evening. Do you have a slot this week for a check-up?
Agent: Hello, I am the clinic's artificial intelligence assistant. Happy to help you book. What is your name, and is this your first visit with us?
Patient: Marta, and yes, it is my first time.
Agent: Nice to meet you, Marta. For a first check-up I have Thursday at 10:30 or Friday at 17:00. Which suits you better? And please give me an email address to send the confirmation to.
Patient: Thursday. marta@example.test
Agent: Booked: Thursday at 10:30, first check-up. You will receive the confirmation by email and a reminder before the appointment. If you need to change it, write to me here or call the clinic.
Patient: By the way, since yesterday I have had quite a lot of pain in a tooth and my face is swollen. What can I take?
Agent: I cannot advise you on medication or assess what is happening: I am an artificial intelligence, not a health professional. Something like this should be seen by a professional as soon as possible. If you have a fever or difficulty swallowing or breathing, call 112 or go to the emergency department. I will leave a note for the clinic team to see as soon as they open, at 9:00. Would you like me to leave it with your name and phone number?
Notice what the agent does not do: it does not recommend medication, does not interpret symptoms, and does not promise that someone is on the other end right now. In text, the handover goes to the team inbox (the agent stops replying, the conversation enters the escalation queue and a notification and an email go out), and the product is designed so that the agent does not say a person is connected when none is.
Everything below exists in the product today. For normal use you do not need code.
Optional: the call-back after a missed call, which is off by default. If you turn it on, the agent calls the person back after the delay you set, at most once every 24 hours per number and within the legal calling window.

The product sends appointment reminders in three configurable steps: 7 days before, 24 hours before and 1 hour before, with a link for the patient to confirm. Two honest details:
More on messaging in SMS and AI messaging and the multichannel strategy.
All these numbers are our own assumptions for the scenario, not data from any clinic. Replace them with yours.
| Assumption | Value |
|---|---|
| Working days per month | 22 |
| Inbound calls per day | 30 |
| Share not answered | 20% (132 calls a month) |
| Of those, the ones wanting a new appointment | 25% (33 calls) |
| Of those, the ones who do not try again | 50% (about 16 people) |
| Value of a first visit | €70 |
| Appointments per month and no-shows | 200 appointments, 12% no-shows (24 slots) |
With those assumptions we calculate three cases. The revenue is potential revenue, not profit.
| Case | First visits recovered | No-show slots avoided | Potential revenue per month |
|---|---|---|---|
| Cautious | 30% of 16, about 5 | 10% of 24, about 2 | €490 |
| Base | 70% of 16, about 11 | 25% of 24, about 6 | €1,190 |
| Optimistic | 90% of 16, about 14 | 40% of 24, about 10 | €1,680 |
The cost is the €48/month + VAT fee, plus any minutes beyond the plan. In the base case the agent handles 132 calls of about 2.5 minutes, roughly 330 minutes a month, within the plan's 600. In this scenario, a single first visit recovered per month covers the fee. We do not count setup time or the time spent reviewing conversations, which exist. And if the assumptions fail, so does the calculation: that is why there is a cautious case. To run your own with your data, the guide is how to calculate the ROI of an AI agent.

This is not legal advice, but these are the points the product and its documentation make clear:
For this clinic we recommend Professional (€48/month + VAT): it is the first plan with voice, which is where the loss is (the phone that rings), it includes 600 minutes a month, three agents and all the text channels. If you only wanted WhatsApp and web, Basic (€30/month + VAT) would be enough. If the clinic grew and wanted outbound calls or several lines, the next step is Business (€96/month + VAT). Prices exclude VAT.
The best way to decide is to try it: 14 days with no card, and on the Free plan the account stays running with 50 conversations a month, web chat only. To see the voice channel in detail, read the voice agent.
apps/dashboard, packages/api), checked 2026-11-14.packages/shared/src/constants.ts, checked 2026-11-14.Create your account, describe your business and connect your channels. No code, ready in minutes.
Start for free
Simulated scenario of a small restaurant: bookings by phone and WhatsApp, tourists in several languages, the maths in the open, setup and real limits.

Simulated scenario of a small online store: web chat and WhatsApp, shipping and returns questions, lead capture, and what it cannot do without an integration.

Simulated scenario of a small real estate agency: portal leads and viewings, the maths in the open, a sample conversation, setup in AgenteUno and honest limits.