1. It says it is an AI
The assistant identifies itself as the clinic's AI assistant in the first sentence of every call, before anything else happens. It does not wait to be asked and it never claims to be a person.
The disclosure is part of the greeting, so a caller hears it whether or not they stay on the line. This is what Article 50 of the EU AI Act requires of a system that interacts directly with people.
2. What it does
Within a clinic's phone line, the assistant handles the scheduling side of a call:
- answers when reception cannot, including outside opening hours;
- offers available appointment times and books one;
- moves or cancels an existing appointment;
- confirms the booking to the caller and writes it into the clinic's calendar;
- speaks the caller's language, and switches language mid-conversation if they do.
3. What it will not do
The assistant is built to stop rather than improvise. It does not give clinical advice, interpret symptoms, discuss medication, or read out test results. Anything of that kind goes to a clinician.
It does not diagnose, it does not triage, and it does not make decisions about a patient's care. It handles scheduling and hands over everything else.
State plainly whether the assistant can take payment details, and whether it can access a patient's medical record in order to identify them. Both are questions a cautious buyer asks early, and both change how the system is classified under the AI Act.
4. Reaching a person
A caller can ask for a person at any moment and does not have to explain why. The assistant transfers the call or arranges a callback, depending on what the clinic has set up. It does not argue, and it does not try to keep the caller on the line.
Some situations escalate on their own, without the caller having to ask: clinical questions, complaints, and anything that sounds like an emergency.
Describe what happens when no human is available, for example at three in the morning. Does the assistant take a message, offer an emergency number, or stay on the line? This is the case regulators and clinics both ask about.
5. Recordings and transcripts
Before a clinic goes live, it receives a plain list of what is captured on a call, where it is stored, who can see it, and for how long. Fields the clinic does not want captured can be switched off before launch.
- Is call audio recorded by default, or only the transcript?
- Is the caller told about recording at the start of the call, in the same breath as the AI disclosure?
- Default retention period for audio and for transcripts
- Whether call content is used to train or improve models. If it is not, say so here: it is a question every clinic asks.
6. Who is responsible for what
When the assistant answers a clinic's line, the clinic decides why and how patient data is used, and Wibe Labs acts on the clinic's instructions. In data protection terms the clinic is the controller and Wibe Labs is the processor. That relationship is set out in a data processing agreement signed before launch.
For the enquiry form on this website the roles are different: there Wibe Labs is the controller. See the privacy notice.
7. Oversight and mistakes
The assistant is a tool operated by the clinic, not an autonomous system. A person can review what it did on any call, and the clinic can change or switch it off at any time.
State how a clinic reports a problem with a call and how quickly it is looked at, and give a contact address for questions about the assistant's behaviour. Without a named route this section reads as a promise with nowhere to send it.