What the assistant is
Each business sets up its own assistant. It uses a large language model from a provider such as OpenAI, Anthropic, Google, Groq, OpenRouter or Cloudflare to write replies. The business decides its instructions, tone and what it knows. A notice in the chat window tells you that you are talking to AI.
How it decides what to say
- When you ask a question, we search the business’s own knowledge (its web pages, documents and hand-written answers) for the most relevant parts.
- Those parts, the business’s instructions and the conversation so far are sent to the AI model, which writes a reply.
- If the knowledge does not cover your question well, the assistant says so, offers to pass you to the team or asks for your contact details instead of guessing.
- For some questions, the business may let the assistant look things up for you, such as an order status or free appointment times.
What it can get wrong
AI can misunderstand a question or state something that is not true. Please check anything important, such as prices, availability, medical, legal or financial information, with the business directly. The assistant cannot make binding promises unless the business has confirmed them.
What happens to conversations
- The business can read the conversation in its inbox and use any details you share to reply to you.
- The message is sent to the AI provider only to generate the reply.
- The business decides how long conversations are kept. You can ask it to see or delete your data; many businesses using PoriFyx Chat can do this in a few clicks.
Model training
We do not use conversations to train AI models, and we use our providers’ business APIs, which do not train on API data by default. We may look at small, anonymised samples only to fix problems with the service.
Humans stay in charge
You can always ask to talk to a person. The business is notified and can take over the chat; once a person has taken over, the AI stops replying. The assistant does not make decisions with legal or similarly significant effects about you.
EU AI Act and similar laws
Chat assistants built on PoriFyx Chat are designed as limited-risk AI systems. We meet the transparency duty by clearly telling people they are interacting with AI, and our acceptable use policy forbids high-risk or prohibited uses.
Questions
For questions about a specific assistant, contact the business that runs it. For questions about how PoriFyx Chat works, email [email protected].