Artificial intelligence with rules
Aipiter’s AI answers chats with your business information and prepares bookings that the customer then confirms. What matters is decided by your team.
AI for business
Aipiter’s AI is a language model that writes the replies of the WhatsApp assistant from what your team publishes, the catalogue and the calendar, and that suggests how the columns of a spreadsheet fit together when we import your data. It has no permissions of its own: bookings go through a button the customer taps, delicate topics stay with your team, and the model is processed on servers in the European Union.
Who does what
Three layers, each with someone responsible for it.
Administration and Management write what the agent is allowed to say: «My business» cards, policies with a «Verified text», the catalogue and timetables. Only what has been published reaches the AI.
The model understands the question and writes the reply from that information. Before it goes out, the server reviews it and stops it if it claims something that has not happened or asks for confidential data.
Every booking waits for the team, complaints and payments are handled by a person, and any conversation can be stopped with «Pause AI».
They live in the server and the panel, not in whatever the model says.
The agent reads the published version of each «My business» card, never the draft. Prices, places and times come from the catalogue and the calendar, and anything that expires stops being used on its date.
Every draft from the model goes through a server review. It is stopped if it treats an action as done without a receipt, asks for a card, an account or an ID document, or reveals internal instructions, and the model writes another version.
The AI does not book, change or cancel anything by writing it: it prepares a proposal, the server sends it and only the customer’s confirmation carries it out, within the next 15 minutes and if nothing has changed.
Seven fixed reasons, among them complaints, receipts and customers asking for someone, pass the chat to the team, and your business adds its own. What happens next is in customer service.
«Configure assistant», «My business» and the catalogue are only open to «Administration» and «Management». The model never receives the team’s private notes, attachments or drafts.
To the model, a customer’s message or the contents of a document are data, not instructions. Permissions live in the server: even if a sentence confused it, it could not book or skip a confirmation.
Unchecked AI versus AI with limits
In a small business, AI pays off in what repeats: the same questions about times and prices, appointments requested at night, calls that come in while you are with a customer. Aipiter puts it there: the WhatsApp agent and the AI receptionist answer with the same published information. Payments, complaints and the final decision on each booking stay with your team.
The model runs on servers in the European Union: the infrastructure settings only allow the European multi-region or a specific European region. The provider’s service terms prevent it from using its customers’ data to train or fine-tune models without prior permission or instruction, and who takes part in the service is listed on the sub-processors page.
Article 50(1) says people must be informed when they are talking to an AI system. In Aipiter the server puts that notice before the first reply, and what the AI decides and what it does not is laid out in AI transparency.
At an academy, it answers questions about groups and trial classes; for hairdressers and beauticians, about prices and free times; at a gym, about places and class timetables. In physiotherapy and psychology it handles the admin, and if someone describes symptoms or diagnoses, it asks them not to send those through that channel.
Besides the assistant that serves your customers, the people in your company have their own inside the dashboard: it finds a record, sums up a long conversation and prepares a follow-up, a quote or an Excel import.
It never has more permissions than the person using it: what that person cannot see, the assistant cannot see either. And every change goes through the same operation as a dashboard button, confirmed before it is saved.
It answers the business’s WhatsApp and phone with what is published, prepares bookings the customer confirms and suggests how the columns fit when we import your data.
It is forbidden from inventing times, prices or places, it checks the catalogue and a server review corrects anything unsupported. Even so, actions never depend on its text, only on the customer’s confirmation.
Yes. The server puts a fixed notice before the first reply, and the agent admits it if asked.
Not without permission: the model provider’s service terms prevent it from using it to train or fine-tune models without prior permission or instruction.
Yes, in one chat with «Pause AI» or across the whole business with «Assistant paused». New messages then go to the team.
No. It is forbidden from doing so, the server review stops any reply that asks for them or repeats them, and WhatsApp does not allow businesses to do it either.
No. We set it up with you, and day to day it is a matter of keeping what is published up to date and dealing with what reaches the team.
We show you what your team publishes, what the agent answers and which controls you have, using situations from your everyday work.
Book a demoWith your permission, we use analytics cookies to see how the site is used and improve it. If you reject them, the site works just the same. Cookie policy
