When you click a conversation in the inbox, the center pane shows the full transcript and the right pane shows context cards. This article explains every element on those two panes.
In this article
The session header
The transcript
AI messages: sources, products, comparisons, quick replies
Customer messages and feedback
The Conversation data card
The User data card
Reading the conversation: a checklist
The session header
At the top of the center pane sits the Session actions card with:
Title — the auto-generated session title, e.g. "Where is my order?".
Hand over / Give back to AI — toggles human takeover (see Taking Over from the AI).
Resolve — marks the session Resolved. Available on Open sessions only.
If the session has already been handed over to a human, an info banner at the top reads: "The chat hand-overed to you. Feel free to share your own suggestions or ideas beyond the AI's input."
The transcript
Below the header, the transcript scrolls bottom-up (latest message visible). Each message is one of three roles:
User — what the shopper said. Plain bubble.
Assistant — what the AI said. Rich bubble that can include rendered text, product carousels, product comparisons, and quick replies.
Human — what you said after handing over. Plain bubble on the assistant side, attributed to a human responder.
AI messages: sources, products, comparisons, quick replies
AI responses are built from typed "sections" so a single message can render multiple things in order:
Text — the answer itself. Markdown is rendered, links are clickable.
Product carousel — a row of product cards (suggested products) with images, names, prices, and "Only N left!" / "Out of stock" tags where applicable. Each card has a tagline reading "Here is a list of related products."
Product comparison — a side-by-side comparison table when the shopper asked Convi to compare items.
Quick replies — chip buttons under the message that suggest the shopper's likely next question.
Each AI message also has a Sources tooltip that reveals two lists:
Knowledges — the custom FAQs, pages, policies, or imported documents the answer was drawn from.
Abilities — the abilities the assistant used to produce the message (e.g. Order tracking, Web search).
This is your audit trail. If a shopper later questions an answer, the Sources tooltip tells you exactly where it came from.
Customer messages and feedback
If the shopper left a thumbs-up or thumbs-down on a specific AI message, a small badge appears under that message:
Positive feedback by user — green, thumbs-up icon.
Negative feedback by user — amber, thumbs-down icon.
Per-conversation feedback (the overall thumbs the shopper leaves at the end) shows up on the right pane.
The Conversation data card
The top card on the right shows the conversation's metadata. It is collapsible.
Summary — a short auto-generated synopsis of the chat.
Topics — one or more topic chips (e.g. Shipping, Returns).
Started chat from — the page URL the shopper opened the widget from. The shop's own domain is stripped for readability.
The User data card
The bottom card on the right shows who the shopper is. Also collapsible.
Name — falls back to "Anonymous User" if the shopper didn't introduce themselves.
Email — the email the shopper provided, or
—if none.Phone —
—(Convi doesn't ask for phone today; the field is reserved).Local time — the shopper's current local time, based on their timezone.
Location — City, Country if both are known; one or the other otherwise; "Unknown" if neither.
User's satisfaction — Positive / Negative / Unknown. If the shopper left a negative free-text comment, an info icon appears next to the value with the comment in a tooltip.
Reading the conversation: a checklist
When you open a conversation to review or audit it, this is the fastest path:
Right pane → Conversation data → Summary for the headline.
Right pane → Topics to see what categories Convi tagged it under.
Center pane for the full back-and-forth.
On any AI message that surprised you, open the Sources tooltip to see what knowledge Convi drew from.
If the answer was wrong, Train → Content is the place to fix the underlying source.