Every customer message becomes a ticket. The AI tries to answer it first from your knowledge base; if it can’t, the ticket is handed to a person on your team — with the whole conversation and a suggested reply ready to go.
A ticket’s life
Every ticket sits in one of three behavior categories. The status names you see (New, Waiting on customer, Resolved…) are set per desk, but each name always belongs to one of these three categories — and it’s the category, not the name, that drives the queue:
| Category | What it means | Example status names |
|---|---|---|
| Open | Needs work — the AI or an agent still has to act. | New, In progress |
| Pending | Waiting on the customer to reply; nothing for you to do until they do. | Waiting on customer |
| Closed | Done. Closed tickets drop off the open list. | Resolved, Closed |
Handling a ticket in Teams
Every new inquiry posts an adaptive card to your support channel. You can reply right from the card, or open the ticket in the AB Support tab to see the full history, the AI’s suggested reply, and the customer’s past tickets. To scan and triage everything at once, work the ticket queue in the tab.
- Read the card. A new inquiry posts to your channel with the customer, their question, and whether the AI already answered or it needs a human.
- Reply from the card. Type your reply and press Send reply — it reaches the customer on the channel they came in on (live in the chat widget, or by email).
- Or open it in the tab. Use Open in AB Support to work the ticket in full — reply, assign it to a teammate, and mark it resolved.
Internal notes — for your team only
A ticket has a separate notes panel for your team. Notes are never sent to the customer — they’re for context, handoffs between agents, and the “watch out for this customer” heads-up. Use them for anything the next agent should know but the customer shouldn’t see.
The activity timeline
Each ticket keeps an activity timeline — an internal-only history of what happened to the ticket: when it was created, status changes, assignee changes, and any undelivered replies. It’s kept separate from the customer conversation, so you can see the ticket’s handling history without it cluttering the actual back-and-forth with the customer.
Resolving and reopening
When you resolve or reopen a ticket, you pick a short reason. There are quick-picks — Answered Duplicate No response Out of scope — or you can type your own. The reason is recorded on the activity timeline, so later you can see not just that a ticket closed but why.
When a reply can’t be delivered
Sometimes a reply can’t reach the customer — for example a WhatsApp message sent outside the 24-hour window WhatsApp allows for replies, or a customer who has opted out. When that happens, the undelivered reply is logged on the timeline instead of failing silently, so an agent knows to follow up another way rather than assuming the message went through.
The more the AI can answer on its own, the less lands on your team — so keep its knowledge sources pointed at good, current content.