So your team can be
where it matters most.

Yes, but less than you think. You get a simple dashboard that shows which requests Ella has taken and which are passed to you, sorted by type of task. You no longer jump back and forth between all sorts of calls. You can work through ten prescription requests in one go, for example. It takes some getting used to, but for most people it is a relief: less switching between things in your head.

Honest answer: it can feel that way. There are visibly more tasks on the table, because nothing gets lost on the answering machine or forgotten. Realistically, the total amount of work stays about the same, it just becomes more visible and better grouped. Most people find after three weeks that they are less worn out, because the phone is no longer interrupting them all the time. And you lose fewer patients and customers, which saves you bigger problems down the line.

No. The dashboard shows aggregated figures (how many calls, which topics come up often, average handling time), never a judgement of individual people. Individual call recordings are only made for quality assurance, and only if your practice has agreed to it. You are not being assessed. You are using a tool.

Ella is deliberately built to say “I do not know that” rather than guess. When that happens, the call is passed to your team along with a short summary of what has been said so far. That way you do not start from scratch: you know straight away what the call is about. And if the AI does get something wrong at some point, that is logged in the system, reviewed by us and put right in the next update. We are liable for what our systems say.

Honestly? In our experience, no. What you do in person, reading people, picking up on worries, weighing things up, carrying responsibility, no AI does that. What Ella takes on is what you are overqualified for: routine calls, repeat questions, appointment coordination. If anything, the demographics point the other way: the EU is short of 1.2 million trained staff. You are needed, just somewhere other than the phone.

Yes. There are modes: “Ella only answers when I cannot get to the phone”, “Ella answers everything and I get the summaries”, “peak times only, Ella steps in when it gets busy”. You set this up yourselves, and you can change it from day to day if you need to.

We do. Onboarding happens online (or in person, if you prefer) and we take the time your team needs. It is not a lecture, more sitting side by side: you show us how you work, we show you how the system fits in. After that you get a few days of hands-on support and a phone number for any questions. No one is thrown in at the deep end.

You are speaking to a voice that helps out on the phone at the pharmacy or MVZ. We call her Ella. She is not a person. She can deal with many requests directly, and for anything that is not clear-cut she passes it on to a human. You can also ask for a person at any time, just say “I would like to speak to someone”, and you will be put through.

Of course. Ella is only the voice on the phone that steps in when no one can pick up, because the queue at the counter is long, because the team is with patients in person, because someone rings outside opening hours. The pharmacist, the pharmacy technician, the medical and nursing teams, they are still your people to talk to. Ella takes pressure off them so they have time for you.

Your details are processed in line with GDPR, and only for the purpose of your request. What you say stays between you, Ella and your pharmacy or practice. We at Ella Media do not use the content of your calls for anything else. You will find the technical details on processing and data flow in the privacy policy.

Only if your pharmacy or MVZ has set this up, and you will be told at the start of the call. By default, only what the call was about is kept as a short note (for example “Ms Schmidt is asking about her prescription for medication X”), with no voice recording.

Yes. In a medical emergency, Ella points you straight to 112 or the medical on-call service on 116 117. For requests that are urgent but not acute, a prescription for tomorrow, an appointment you need soon, she puts you straight through to a person. She never just hangs up because you are upset.

Just say so. Something like “I would like to speak to someone” is enough. We take your scepticism seriously, and we do not assume everyone is happy talking to an AI. Some are, some are not. Both are fine.

We built the system for exactly this, trained on a wide range of voices. Older voices, regional accents, the slower speech of many older people: that is the norm, not the exception. With very pronounced speech difficulties or unusual accents, Ella notices when she is not sure she has understood, and hands over to a person sooner. We do not want anyone who used to manage phone calls to be left behind.

Ella processes data in line with GDPR. Audio is processed through EU data processors under the appropriate data processing agreements. The AI components we use are documented and curated for the medical and pharmaceutical context. For specific architecture details (model providers, data flow, region), we provide a technical overview on request during the initial call.

Standard connections to the common practice management systems (appointments, patient master data, medication plans) and to pharmacy inventory management (availability, prescription identifiers). Depending on the provider, interfaces can be API-based, via HL7/FHIR or via structured export formats. For specific or older systems, we assess them case by case and price the work transparently.

Every system access and every request is logged with a timestamp, request type and result. Your data protection officers can view the log. It is GDPR-compliant and, if required, configured to account for §203 StGB (the duty of confidentiality). We are happy to show you the audit log specification before you sign.

Depending on the configuration, our systems fall between “limited risk” (standard cases) and “high risk” (where medical decisions are being prepared). We run our models with full, traceable documentation, human-in-the-loop escalation and a declaration of conformity. For more critical applications, we produce a separate risk assessment before go-live.

If the connection drops briefly, calls are forwarded under our fallback rule (for example to your answering machine, or diverted to your team). We agree your specific fallback setup in the initial call. In a power outage, your normal phone emergency arrangement takes over.

Domain-adapted language models (LLMs), curated for the medical and pharmaceutical context, combined with Retrieval-Augmented Generation (RAG) over your own knowledge base (active ingredients, indications, in-house rules). Before every answer, the system checks whether the model is confident enough; if not, it escalates. We document the components we use transparently in the initial call.

No. We manage your setup remotely, roll out security updates and model improvements automatically (and let you know), and handle patch management entirely. You can pause or roll back updates at any time. For CISO requirements, we provide a full SBOM (Software Bill of Materials) on request.