AI Assistant Notice
AI Assistant Notice
1. You are talking to an AI system
The chat assistant on the xeli website (https://xeli.pl) is an artificial intelligence (AI) system. We inform you of this in line with Article 50 of Regulation (EU) 2024/1689 on artificial intelligence (the AI Act). You see this information from the very first message in the chat. The Assistant is provided by Gryf.ai Dawid Wachałowicz (“we”, “us”).
2. How the Assistant works
The Assistant uses three sources of answers. Every message is labelled with the source it comes from:
- “Approved answer”: a ready-made answer that we have prepared and approved in advance. Prices, deadlines and terms of cooperation come from such answers.
- “AI answer”: an answer written automatically by a language model on the basis of information about our offer and the content of the Website. It appears when a question does not match any script.
- “Consultant”: an answer written personally by a person from our company.
We also mark AI-written content in the page code, in a machine-readable form. The labels are also kept in a conversation transcript sent by e-mail.
When AI answers are switched off or their daily limit has been used up, the Assistant replies with scripts only and suggests the contact form.
3. What AI can and cannot do
- AI can make mistakes and give incomplete or outdated information. Check important information in the price list on the Website or ask a Consultant.
- Prices and deadlines are given in answers labelled “Approved answer” and in the price list. If an answer labelled “AI answer” gives a price or deadline, treat it as indicative. We confirm the final terms in an individual offer.
- AI does not conclude contracts, accept orders or make binding promises.
- AI answers are not legal, tax or financial advice.
- The Assistant can help you on the Website: scroll to a section, highlight an element, start a demo or fill in a form field. You see changes in the form immediately and can undo them. You always send the form yourself, with your own click.
- AI prepares the whole personalised demo: the texts, the colours, the choice of tool, the common questions with their answers and the replies of the assistant on the demo page. The demo and the results of other demos are for illustration only and may contain mistakes.
- We do not use AI to recognise emotions or to identify people by their voice or appearance.
- The tool demos on the Website have chats of example businesses, such as a salon's reception. Ready-made answers reply first, and a question they cannot answer goes to an AI model through OpenRouter. AI answers carry the same label as in the Website chat. You can see the conversation live in the demo's example panel. We do not store it as a conversation with the Assistant and delete it after 24 hours. We limit the number of AI answers per visit and per IP address.
4. When a human replies
- A Consultant can take over the conversation at any time. From then on, their messages are labelled “Consultant”.
- A Consultant may read conversations live so that they can help quickly when needed.
- If you prefer to talk to a human, say so in the chat. You can also send the contact form, write to kontakt@gryf.ai or call +48 726 001 488.
- When none of us is available, the Assistant will ask you to get in touch through the form. We will reply as soon as possible.
5. Suggestions and dictation
- The Assistant may offer help on its own, for example when you spend a while looking at the price list. Such messages are ready-made scripts. They appear rarely and never while you are typing. You can mute them for the rest of your visit.
- You can dictate a message. The text appears in the message field, so you can check and correct it before sending. Depending on the settings, speech is converted to text by your browser or by an AI transcription model we have selected.
6. Your data in the conversation
- We send the AI model your message with any photos or PDF files you sent, the latest part of the conversation (older messages as a summary), the page context (the section you are viewing, the language, data entered in our forms) and information about our offer. When a conversation relates to your inquiry, AI prepares a short summary of it for our team.
- AI models are made available to us by OpenRouter, Inc. (USA). A request may reach a model provider outside the European Economic Area. We ask for requests to be routed to providers that do not train models on the data and do not retain it, where this is possible.
- We keep the content of conversations for 12 months in order to answer questions and improve our offer. We do not use conversations to train AI models.
- The conversation is saved in the browser's session storage only when you open the chat, and it disappears when you close the tab.
- Do not share passwords, payment card details, health data or other specially protected data in the chat.
- Details, including your rights, are set out in the Privacy Policy.
7. Who is responsible for the Assistant
The Assistant has been prepared and is provided by us: Gryf.ai Dawid Wachałowicz, ul. Wspólna 26a, 37-420 Rudnik nad Sanem. We use AI models from other companies. We decide on the ready-made answers, the rules the Assistant follows and the choice of models ourselves.
8. Reporting errors
You can rate each answer in the chat. If the Assistant gave incorrect or inappropriate information, write to kontakt@gryf.ai. We check every report and correct wrong answers in the scripts. We handle complaints in accordance with the Terms.
This English version is a translation; in case of any discrepancy, the Polish version prevails.