In businesses of five to thirty people, availability is rarely a matter of concept. It is a matter of simultaneity: the phone rings exactly when the responsible person is with a customer, in the workshop, or on another call.
The usual answers are both unsatisfying. Many callers will not use voicemail; they hang up and call the next supplier. A dedicated receptionist only pays off above a call volume many businesses do not have. Between the two sit several steps that rarely get thought through.
The steps between voicemail and a dedicated role
- 1.Routing rules: by time, by request, in sequence across several numbers. Costs nothing but thought and solves surprisingly much.
- 2.External call answering: a service picks up and sends a note. Reliable, but without knowledge of your business.
- 3.Automated answering with request capture: picks up, asks for the request and a callback number, and creates a case from it.
- 4.Combination: people first, automation only outside the hours when someone can pick up.
The fourth variant is the one best received in daily use, because no customer ends up in an automated system while someone is sitting in the office. It does require that the hours are recorded realistically rather than aspirationally.
Where it fails in practice
Not on the technology. Almost always on what happens after the call.
- The notes land in a mailbox that belongs to nobody in particular.
- There is no agreed time within which people call back.
- There is no overview of open callbacks, so nobody notices when one is dropped.
Automated answering without a settled callback route makes things worse, because the customer now expects someone to get in touch. Those three points therefore belong before the technology, not after it.
What is added from August
On 2 August 2026 the transparency obligations of the AI Act become applicable. Anyone running automated answering with an AI system should have the announcement worded by then so that it is clear what the caller is talking to and how to reach a person. In our experience that is no disadvantage: callers react more calmly when they are told than when they work it out.
How does AI call handling work in a medical practice?
In a medical practice the load is rarely spread across the morning; it lands between eight and ten. That is exactly when the phone rings while someone is standing at the desk. AI call handling answers every call during that window, records the name, callback number and concern, and turns it into an entry the team works through once there is room.
- Answer calls outside consultation hours instead of leaving them in a queue or unanswered.
- Take appointment requests and match them against open slots, with confirmation by email or text message.
- Pass prescription and result enquiries to the team as a task, with everything needed to act on it.
- Put callers who say the matter is urgent straight through to a stored number.
AI call handling gives no medical advice. It records, sorts and forwards. Anything that needs judgement goes to a person.
Health data carries special protection. We therefore agree in advance which details are recorded at all, where they are stored and for how long. We set out a data processing agreement and storage inside the EU for this, as far as the provider in use supports it. Our page on GDPR and the AI Act explains what that means in detail.
We first look at how many calls actually go unanswered, and when. Often the first step is then enough, and the topic is settled without introducing a system at all.