Home care on-call rota, organised — without burning out the PDL
In many home care services the on-call phone rests on two or three people — often the PDL (the nursing service manager), who still runs the 7 a.m. handover after a night on the phone. An AI phone assistant cannot replace on-call cover, but it can pre-filter it: the on-call phone rings mainly for what you have described as urgent.
Why on-call always lands on the same people
On-call cover in a home care service needs people who can judge what is happening at the other end of the line — and there are too few of them. 36 % of companies cannot fill open positions, at least in part (DIHK skilled-labour report 2025/26), and according to the German Federal Employment Agency, recruiting new staff for care is still harder than average. So on-call rotates within a small circle, or it simply stays with the PDL (the nursing service manager). Every manager knows the result: broken nights because of calls that could have waited until morning.
The common models — briefly and honestly
A rota across the team: fairly shared, but every qualified member of staff loses nights on a regular basis — and not everyone is allowed to carry that responsibility, or wants to. Everything on the PDL: works until the PDL is out of action. An external emergency call centre: takes the load off, but knows neither the rounds nor the clients and passes a lot on unfiltered. AI as a pre-filter in front of it: changes nothing about who is on call — but a great deal about how often they are woken.
AI as a pre-filter: on-call is woken less often
The principle: calls outside office hours are answered first by the assistant — immediately, with no queue. It asks calmly and in a structured way what the call is about, and applies your rules. A fall, an acute deterioration, a door that does not open for a scheduled visit: when it recognises a request that your rules describe as urgent, it puts the call straight through to the on-call number you have stored; if nobody answers there, it takes an urgent message. The question about the round for the day after tomorrow, the prescription query, a relative's organisational question: those it takes down as a message, with the name, the request and the callback number (a dictated number it reads back digit by digit) — and in the morning it is sitting sorted in the team inbox.
What counts as "put through immediately" is something you set in the rules, for example: concern "fall or acute deterioration" → straight to the on-call number. That does not turn the assistant into an emergency call centre: scoco.ai is not an emergency service and does not guarantee that anyone is reached. The recognition is a help without guarantee (AGB § 2 (4)): an urgent call, too, can end up as a message in the inbox. The "Another industry" industry template, which scoco.ai proposes for a home care service, has no built-in pointer to 112 (there is no dedicated care template). So put the pointer to 112 into your greeting or knowledge base — the AGB oblige you to refer callers in urgent cases to suitable services and to check notifications at reasonable intervals (§ 7 (4a)) —, check notifications at night too, and tell clients and relatives in advance, for instance at the initial assessment, that in an acute medical emergency 112 comes first. Questions the assistant answers from your knowledge base, and it is instructed not to guess. Unlike the practice templates, "Another industry" contains no instruction not to give medical assessments.
Relatives call at night because they are worried
At night it is not only clients who call but relatives too — often worried, sometimes upset, by no means always with a genuine emergency. For whoever is on call these conversations are wearing; for the assistant they are routine: it stays patient, takes the request seriously and in full, and announces the callback. Anyone who would rather not call in German can switch language mid-conversation — the summary for your team still arrives in your account's language, with the transcript translated.
Sensitive data: what happens to the details
A care phone line inevitably takes in health data, and health data belongs to the special categories of personal data (Art. 9 GDPR). Whether and how you use an AI assistant for it is therefore something to clarify in advance with your data protection officer; this article is not legal advice. What scoco.ai supplies as facts for that review: there are no call recordings. Transcripts and summaries are created — which can themselves contain health data — stored on servers in Germany (Falkenstein), with deletion periods you choose yourself. Speech recognition and the language model run in the EU; speech output may run through a US provider that receives only the assistant's reply texts (no caller audio). Which service providers take part in the processing is set out in the sub-processor list; there is a data processing agreement, and for professionals bound by secrecy (§ 203 StGB) it provides a separate confidentiality agreement on request. Access runs through the team inbox, so your team sees every report with its caller and request; what happens after a call in detail is described in "What happens after the call?". For handover into your care management software the honest position is this: the information arrives by inbox and email, and there is currently no direct integration with sector software. If you want call data passed on automatically, you can use webhooks or Zapier, Make or n8n — whether your care software accepts such data is something to settle with its vendor.
Honestly: the professional judgement stays with people
An AI assistant does not triage medically — it sorts by the rules you give it, and in case of doubt those rules should rather have it put the call through. That also means: now and then the on-call phone does ring for something that could have waited — and conversely, an urgent call can end up as a message in the inbox. It does not replace the on-call rota, and it certainly does not replace the familiar conversation with the carer who knows the client. What it does is something more precise: it makes sure the scarce resource "a qualified member of staff on call" is woken mainly for the cases your rules describe as urgent. How the principle works across industries is shown in "Automating the on-call phone".
Frequently asked questions
Is an AI assistant even allowed to take calls involving health data?
There is no blanket answer: health data belongs to the special categories of personal data (Art. 9 GDPR). Clarify the deployment in advance with your data protection officer. For that review scoco.ai provides a data processing agreement and the sub-processor list; there are no call recordings, you choose the deletion periods yourself, and speech output may run through a US provider that receives only the assistant's reply texts.
How does the AI know what counts as a care emergency?
From your rules: you describe which requests are put straight through to on-call; everything else the assistant takes as a message. It matches the request against those descriptions, without guarantee (AGB § 2 (4)): an urgent call, too, can end up as a message. That is no substitute for a medical judgement, and its template has no built-in pointer to 112. So put one into the greeting or knowledge base (AGB § 7 (4a)), and tell clients and relatives in advance that in an acute emergency 112 comes first.
What if relatives do not want to speak to an AI?
It happens. The assistant says openly that an AI is speaking — that is required. What matters is what happens next: the request is taken down, the callback comes reliably. If you have set up a rule for urgent requests, it puts the calls it matches through to the on-call staff.
Does this make the on-call night shift unnecessary?
No — you still need a qualified member of staff who can be reached. They are simply woken less often, because requests that none of your rules describes as urgent do not land with them, but instead sit sorted in the inbox in the morning.
Status & sources
As of 2026-11-15.
- DIHK skilled-labour report 2025/2026, published 19 December 2025 (nearly 22,000 companies): 36 % cannot fill open positions, at least in part
- German Federal Employment Agency (Bundesagentur für Arbeit), press release no. 21 of 9 May 2025 ("Neue Pflegekräfte kommen vor allem aus dem Ausland"): recruiting new staff for care is still harder than average
- EU AI Act Art. 50 (disclosure duty)
- GDPR Art. 9 (special categories of personal data, including health data)
- scoco.ai product information: as of 15 November 2026
- scoco.ai data processing agreement, as of 15 November 2026, § 2 (3) (special categories) and § 4 (4) (separate confidentiality agreement for professionals bound by secrecy, on request)
- scoco.ai general terms and conditions (AGB, German only), as of 15 November 2026, § 2 (4) (not an emergency or on-call service; classifying a call as urgent is a help without guarantee) and § 7 (4a) (the customer sets the service up so that callers in urgent cases are referred to suitable services, such as 112, and checks notifications at reasonable intervals)