Wijkzorg Vlietstad, a fictional example.
Wijkzorg Vlietstad is a fictional organisation (home care & community nursing) that Zorgvorm built to show how such a website works, for visitors and for the team. The organisation, its people and its data do not exist; the website really works.
A complete website for a home-care organisation, from requesting care and an honest question assistant to applying for a job in two minutes. On the team side every request arrives already summarised, with a call script, missing details and suggested intake times.
The problem behind this example.
At a home-care organisation almost every request for care starts on the phone. A daughter calls for her father, does not know which neighbourhood team covers his street and has half the details to hand. The coordinator takes notes, calls back for what is missing and answers the same questions about funding and waiting times in between. On top of that, the website has to recruit nurses and carers.
Two sides of the same website.
What a visitor does arrives on the team side. Try both; all data is fictional and stays in your browser.
As a visitor
- Request care in four steps, with the right neighbourhood team
- A question assistant that only answers from approved texts
- Vacancies with a two-minute application
On the team side: Coordinator portal
- Just arrived · summary and call script
- Three intake times suggested
- Unanswered question · add to the assistant
Fixed rules, and where AI helps.
What runs on fixed, explainable rules, and where a language model takes part. Wherever AI is involved, the result says so.
Fixed rules
The request picks the neighbourhood team from the district, sends urgent needs to the phone and puts everything complete in the coordinator’s inbox.
With a language model
The question assistant answers only from approved texts, with source numbers. If the answer is missing, the question becomes a handoff for the team.
Choices for privacy.
- The request asks for practical matters, such as the street, the kind of help and the situation, and never for a diagnosis or citizen service number. The open comments field warns as soon as someone seems to describe something medical.
- The question assistant first runs a privacy filter that blocks recognisable details. Medical questions go to a member of staff without a language model.
- Everything you enter in this example website stays in your own browser. Nothing goes to a real organisation.
- Where a language model is involved, the result says so. It gives no medical advice and does not decide who receives care.
- For a real application we agree beforehand which data is needed, who processes it and under which data processing agreement.
What this means for your organisation.
For a home-care organisation this means a request that arrives complete at the right team, a coordinator who calls back with a summary and call script, and a vacancies page where applying takes two minutes. Which parts you need, and whether the request can reach your care record system, is something we work out together.