1. Tap a card in the list to open it in the center column. 2. Read the thread top to bottom: the **visitor's messages** (or already-identified customer's), the **AI's replies** — marked with the "AI assistant" label — and, if someone has already stepped in, the **team's messages**, marked with "Operator" and the name of whoever wrote them. Every message carries its timestamp. 3. Check the **right panel**: the customer's record with their name, their email if known (or the web visitor's anonymous identifier if they haven't identified themselves yet) and an **Email** button to write to them directly when their address is known. **The AI's interpretations**: next to the customer's name, in the conversation header and also on their record, you'll see a pair of tags — for example, **"Unclassified"** and **"No analysis"** while the AI hasn't drawn any conclusion about that conversation yet. They fill in as the AI interprets the intent and quality of the contact — the same classification that later feeds the **interest** and **quality** columns on the lead (manual 17 · Leads). > **Note — The conversation updates on its own** > > There's no need to reload the page: new messages — from the customer, the AI, or a colleague — show up on > their own in the open thread, and the mode and the unread-message badge on the list to the left update > the same way, in real time. > **Note — The customer already knows they're talking to an AI from the first message** > > For transparency (AI Act, Art. 50), the first turn of every new conversation on WhatsApp, Telegram, web > chat or Messenger prepends a short notice like "I'm a virtual assistant powered by artificial > intelligence…" before the reply — once per conversation, mandatory and not something you or your > administrator can turn off. You'll see it at the start of the thread like any other AI message. Full > detail in **Manual 35 · Training your team on the AI system**.