Introduction
Joris Bente tested a health chatbot at three in the morning, not in an office. His mother had symptoms that were not dramatic enough for certainty and not harmless enough for sleep. The answer was generally useful and, at one crucial point, too vague.
Joris lives in the Uckermark and tests digital health offers from the perspective of family caregivers. He is not a clinician. That boundary is central to his work.
Story of the Path into AI
Care in a rural area often means waiting loops, long drives and decisions made in kitchens. Joris began using health chatbots to prepare questions, but he quickly saw that they had to handle fear, incomplete information and emergency thresholds much better.
He learned health communication, escalation scenarios and user testing for older people. His first project was a collection of realistic caregiver questions in which chatbots must give clear guidance on when urgent help is needed. The early tests exposed a problem: some systems explained a lot before naming the risk.
Joris made his role precise. He does not evaluate diagnoses. He evaluates understandability, timing, escalation and whether a worried person can see the next safe step.
Current Work
Today Joris works with a patient organization on user tests. In one scenario involving shortness of breath, a bot offered long explanations before pointing to acute help. The product team changed the sequence after the test.
Providers now pay more attention to the moment when plain language and quick human assistance matter more than completeness. Joris keeps reminding teams that users do not meet chatbots in ideal conditions. They meet them tired, scared, distracted and sometimes alone.
Personal Advice
“Do not test health AI only in the office. Test it at three in the morning with fear in your stomach,” Joris says. He advises teams to design for the person who cannot read five calm paragraphs before acting.
Key Facts
Age and place: 55, Uckermark.
Background: househusband, family care, rural healthcare access.
Entry into AI: realistic caregiver questions for testing health-chatbot escalation.
Focus today: health chatbots and user safety.
Typical tools: user tests, health communication, escalation scenarios.
Werkstattnotiz
Joris’s test scripts include background noise, tired phrasing and missing information. Clean test questions looked professional and failed the night-time situation. His next round examines whether chatbots can say “seek help now” early enough without turning every uncertainty into panic.