
AI agent for clinics and practices: a simulated scenario, step by step
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 of a small restaurant: bookings by phone and WhatsApp, tourists in several languages, the maths in the open, setup and real limits.
Simulated scenario: this is not a real customer. The restaurant, the people and the numbers in this article are made up so we can do the maths calmly. There is no testimonial and no measured result from any AgenteUno customer. What is real is what AgenteUno can and cannot do, checked in the code.
A family restaurant with 36 seats in a coastal town in Spain: four people between the dining room and the kitchen, open Thursday to Sunday for lunch and dinner, plus a private room for groups of up to 12. In season, half the guests are tourists who speak English, German or French.

The phone rings exactly when everyone is serving. At lunch and at dinner nobody can pick it up; outside opening hours nobody looks at it. WhatsApp messages get answered in the morning. And the questions repeat: is there a table on Saturday, are dogs allowed on the terrace, does the menu have gluten-free options, when does the kitchen close, where can I park.
Everything is our assumption for the scenario, not a measurement. Swap in your own.
| Assumption | Value |
|---|---|
| Calls per month | 180 |
| Those nobody answers (service and closing time) | 30% (54 calls) |
| WhatsApp messages per month | 120 |
| Those received out of hours, unanswered until the next day | 40% (48 messages) |
| Of those unanswered calls and messages, the ones that wanted to book | 50% (51 requests) |
| Requests the agent takes in full (name, phone, date, time, party size) | 80% (41) |
| Of those, the ones that would have been lost entirely without the agent | 30% (12 tables) |
| Average spend per table | 90 € |
| Margin after the cost of product | 65% |
The sum: 12 extra tables a month × 90 € = about 1,080 € of revenue, about 700 € of margin. In a pessimistic version (only 10% would have been lost, 4 tables) you get about 360 € of revenue and about 235 € of margin. At that margin per table, two extra tables a month comfortably cover the plan (€48/month + VAT). We do not count the staff time saved by not interrupting service to pick up the phone: it is real, but it depends on each venue.
Usage: the 54 calls at about 2.5 minutes come to roughly 135 voice minutes, against the 600 in the Professional plan. WhatsApp and web conversations fit easily inside the 2,000 a month.

Saturday, 23:10. A tourist writes to the restaurant's WhatsApp, closed at that hour. Simulated conversation, written by us to show the flow; it is not a real transcript. The agent answers in the language of the message.
Customer: Hello! Do you have a table for 4 next Saturday evening? One of us is coeliac.
Agent: Hello! I am the restaurant's virtual assistant (an artificial intelligence). I can take your request now, and the team will confirm it by WhatsApp tomorrow before 12:00. What time would you like, and what name should I put it under?
Customer: 21:00, under Hansen. My number is the one I am writing from.
Agent: Thank you, Ms Hansen. So: Saturday at 21:00, four people, under Hansen. About the coeliac guest: our allergen sheet lists the dishes marked gluten-free, but for a coeliac guest I will ask the team to confirm with the kitchen, and I will note it in your request. Anything else, such as a birthday or a high chair?
Customer: No, that is all.
Agent: Noted. This is a request, not a confirmed booking yet: the restaurant will confirm it by WhatsApp tomorrow before 12:00.
What happened behind the scenes: the agent did not confirm anything it cannot confirm, it saved the contact, it left the allergy in plain sight, and the conversation, with its summary, waits for the team in the dashboard. That rule ("do not confirm, request") is written in the agent's prompt: it is an instruction, not a lock in the code. Test it with the simulator before opening the line.

The agent answers in the language of the customer's latest message and keeps it if you ask. In chat, the widget itself is translated into 11 languages (including Spanish, English, French, German, Italian, Portuguese, Polish, Russian, Japanese, Chinese and Catalan). On calls, switching language on request is implemented for a list of 19 languages. What the AgenteUno website publishes as guaranteed are Spanish and English; in the others the model usually holds up well, but do not promise it in writing without testing it with real phrases from your customers, especially with allergens and prices.
Professional (€48/month + VAT). It is the first plan with voice, and the phone is a restaurant's booking channel. 600 minutes, 3 assistants, 2,000 conversations a month, 5 seats, one simultaneous line and inbound calls only.
The pattern is the same with different questions: check-in and check-out times, house rules, pets, parking, directions, what is nearby, indicative rates. Two differences are worth knowing. Availability by night and prices by date do not live in AgenteUno: there is no integration with booking or channel managers, and the calendar works in 30-minute slots within a day, not in stays. The agent takes the request (dates, guests, contact) and your team confirms it. And after you confirm, it can send arrival instructions by WhatsApp or email if you upload them to the knowledge base.

It suits you if you lose calls during service, receive messages out of hours, serve customers from several countries and your team can confirm requests every morning. It is an agent that looks after the front door, not a floor-management system.
It does not suit you if you need to confirm tables instantly and automatically with capacity control, if you depend on integration with your booking system without development work, or if you already answer everything that comes in.
Create your account, describe your business and connect your channels. No code, ready in minutes.
Start for free
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 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.