The first point of medical contact is AI — and that's a good thing

AWS Healthcare lead Jens Dommel describes AI as the new first point of contact in healthcare. What does that mean for over 55,000 family-doctor practices in Germany that field 100 to 200 calls a day? Not diagnosis. Not therapy. But this: picking up the phone, booking appointments, organizing callbacks — so doctors can do what they were trained to do.

We take the call. The doctor treats the patient.

What does AWS Healthcare lead Jens Dommel say about AI in healthcare?

Jens Dommel, Head of Healthcare EMEA at Amazon Web Services, put the role of AI in healthcare into perspective in a recent interview. Three key statements:

• Relieving staff is the most urgent priority: "Relieving the burden on staff is one of the most important challenges of the coming years." (elektroniknet.de, January 2023)
• By AWS's own estimate, 97% of medical data goes unused — because it sits unstructured in records, phone calls, and faxes. (AWS/gesundheitswirtschaft.at, 2023)
• Security comes first: "Safety and security are Job Zero for us." (gesundheitswirtschaft.at, September 2023)

What Dommel describes isn't a future scenario. It's a reality that has long since arrived in DACH practices — though at a point that rarely makes headlines: the phone.

What does the reality look like in German medical practices?

The phone is the bottleneck of primary care.

• 5,000 family-doctor positions are unfilled in Germany — and set to double by 2030. One in four family doctors plans to leave the profession within five years. (Bertelsmann Stiftung / University of Marburg)
• 100 to 200 calls a day hit an average practice in the first two hours alone. Averaged across the day, that's 25 to 50 calls, at roughly 3.5 minutes per call — about three hours of daily phone time for the practice team. (mediform.io / agentino.de, 2025)
• 40% of these calls are pure appointment requests. (med2day / KBV accessibility survey)
• Over 90 minutes a day is spent by doctors on documentation — time that's missing from patient contact. (KBV/Medscape via kleiboldt.de, April 2026)

The math is simple: if nearly half of all calls concern appointments and each call takes three and a half minutes, a practice with 50 daily calls burns roughly 70 minutes just on scheduling. Every single day.

What can AI legally do in a practice today?

The question is no longer whether AI belongs in medical practices. The question is: what is it allowed to do — and what isn't it?

The German Medical Association (Bundesärztekammer, BAEK) published a position paper in March 2025 that brings clarity. Symptom checkers in preliminary assessment are described as an "acceptable tool," and triage in emergency departments and video consultations are explicitly permitted — under one condition: responsibility remains with the doctor. According to BAEK, 47% of the population even expect better results from AI than from a doctor in certain cases — but that changes nothing about the legal accountability.

A Swiss study shows that AI-supported initial assessment can work: at Kantonsspital Baden, 2,543 patients were assessed using an AI triage system. Result: zero cases of potentially dangerous under-triage (Meer et al., JMIR 2024). That doesn't mean AI is perfect — it means organizational preliminary assessment under medical supervision can work safely.

What an AI phone assistant can do in a practice:

• Book, reschedule, and cancel appointments — including routing by urgent, preventive, or follow-up care.
• Capture callback requests in a structured way — concern, urgency, preferred time.
• Take repeat-prescription or referral requests — approval stays with the doctor.

This isn't medical advice. This is practice organization.

What is AI explicitly not allowed to do in a practice?

The red line is clear — and it's defined by law.

• No diagnosis. Section 630a of the German Civil Code (BGB): the doctor owes the treatment, not the software.
• No treatment recommendation. The Healthcare Professions Act and the Model Professional Code: diagnosis, indication, and therapy are medical responsibilities and cannot be delegated (BAEK 03/2025).
• No emergency assessment. Even the sentence "Sounds like a cold, can wait until Monday" legally already counts as triage — and falls under high-risk AI per EU AI Act Annex III. The high-risk obligations become binding on 08/02/2026. (CMS Law 2026; Quickbird Medical 2026)
• No medical device. AI that gives medical recommendations is a medical device under the MDR and requires CE certification. A phone assistant that books appointments is not one.

"AI doesn't make decisions, and it must be transparent. Responsibility for care remains with doctors."

— Dr. Sibylle Steiner, board member, KBV PraxisWissen, 05/28/2025

How does an AI phone agent change everyday practice life? Three scenarios

Ben is Amazing Agents' AI phone agent. He handles the first point of contact — not the medicine.

Scenario 1 — Family practice

Monday, 7:58 am. The phone rings before the door even opens. Ben picks up: "Good morning, Dr. Keller's practice. How can I help you?" The patient wants a follow-up appointment for her blood pressure. Ben checks the available slots, confirms Wednesday at 10:15 am, and sends a summary to the practice software. The medical assistant has both hands free for the first patient in the waiting room.

Scenario 2 — Dental practice

A patient calls because their temporary crown is loose. Ben records: "Temporary crown loose, no pain, requests a prompt appointment." That lands as a structured callback request with the practice team — with an urgency level, without any medical assessment.

Scenario 3 — Physiotherapy

Ms. Schneider needs a follow-up prescription for her knee therapy. Ben takes the request: name, insurance number, requested prescription. The approval is made by the therapist — Ben just makes sure nothing gets lost.

In none of these cases does Ben make a diagnosis. In none of these cases does he recommend a treatment. He does what a well-organized front desk does — just without hold music, without a lunch break, and without "all lines are busy."

We take the call. The doctor treats the patient.

What do practices need to watch out for on data protection?

Under GDPR Article 9, health data belongs to the special categories requiring particular protection. For the use of an AI phone assistant, that means:

• A data processing agreement (DPA) is mandatory.
• Patients must be informed that they're speaking with an AI — the transparency obligation under EU AI Act Art. 50.
• Medical confidentiality under Section 203 of the German Criminal Code (StGB) remains unaffected — the phone agent is subject to the same confidentiality obligations as any medical assistant.

Ben is not a medical device under the MDR. He books appointments, captures callbacks, and takes prescription requests. He doesn't diagnose, doesn't assess, doesn't recommend.

Try Ben in your practice

Want to hear how Ben sounds in your practice? Amazing Agents sets up the trial run within two weeks — tailored to your office hours, your specialty, your tone of voice.

No risk. No long-term contract. Just try it.

→ Schedule a demo now