3.1.4
Where is artificial intelligence used?
In short In four places: translation, the job description assistant, detection of contact details and detection of specially protected data. The model suggests or flags; a human or a simple rule decides.
We use the Claude model from Anthropic for four tasks. The model only suggests or flags. A human or a simple rule makes the decision.
- Translating the job description, the comment on an offer and messages. The translation is labelled, the original is shown next to it, and the original is what counts.
- The job description assistant. It asks about details and suggests which work to pick from the catalogue. It introduces itself as AI and publishes nothing on its own: you approve the text.
- Detecting phone numbers and e-mail addresses, which we hide until a professional is chosen. A marker appears where the hidden part was.
- Detecting specially protected data, e.g. about health or religion. We remove it before the text is saved, and you get a notice about it.
What AI doesn’t do
It doesn’t rate professionals, set the order of offers or jobs, or assign jobs. It doesn’t decide on restricting or blocking an account. It doesn’t verify identity or compare faces with documents.
Your data
The model receives only what the task requires. We don’t send the name, e-mail or phone number from your account. We keep conversations with the assistant for 180 days.
Translation can be inaccurate, so if in doubt, ask the other side. Report text hidden by mistake through “Help”. More on the Artificial intelligence in HandyMan24 page.
