Chat, email, SMS, voice and RCS arrive in the same inbox. What reaches a person has already been triaged, summarised and drafted — you are reviewing work, not starting it.
“I’m sorry the lamp arrived damaged. I can refund $48.00 to the card you paid with, and you can keep the lamp.”
The verifier’s note and the AI’s confidence sit on the card. Editing the reply before you send it teaches the next one.
This is the queue doing its job: what the AI can finish on its own, what it wants a person for, and what it hands over entirely.
The lamp arrived cracked. Can I get a refund?
Where is my order?
Out for delivery — arrives Thursday.
I'd like to speak to someone about invoice INV-2291.
“I’m sorry the lamp arrived damaged. I can refund $48.00 to the card you paid with, and you can keep the lamp.”
Example data, in the shape the console uses. Approve & send grants the message and its action together.
The queue on the left, the conversation in the middle, and everything you would otherwise go looking for on the right.
Search, filter by status — all, needs review, AI live, escalated, resolved — and by channel. Each row carries the customer, the lane, the channel, a preview and an SLA countdown, with a coloured rail down the left in the lane’s own colour.
Every message with its author, its time and a day separator, and the AI’s tool calls inline as chips, so you can see what it looked up and when. Live conversations stream in as they happen. An escalated one opens with the handoff packet and a composer in the same pane.
The customer card with their plan, their verification level and the sentiment on this conversation; their open orders; the procedure’s steps with the ones already done and the slots already filled; and the verifier’s report on the draft you are looking at.
Taking a conversation over parks the AI's procedure exactly where it was. Handing it back gives the AI the conversation again — but not the lane.
This distinction matters more than it sounds. If a conversation dropped to Draft because a check failed, it is still in Draft after you hand it back, and the AI still needs somebody to approve what it writes. The lane comes back up only when the rules you wrote say it should, never as a side effect of somebody being helpful.
On a live call the same controls are there in a call bar: watch the transcript as it is spoken, whisper a line only the AI hears, or barge in and take the call. A whisper is carried as a supervisor instruction with your name on it, for exactly one turn, and the reply is checked against it before the AI speaks.
Why it was raisedThe customer asked for a person. Two earlier messages, both answered. The AI had verified the email and found order A‑1042.
Suggested next stepConfirm the delivery address, then reissue.
An escalated conversation opens with its handoff packet: the reason, the summary, what was already tried, and what to do next.
A support tool is only as good as what it does on the thirtieth day. These are in the box.
Internal notes on the thread, and an @mention list that only offers people whose role can open this conversation.
Email a supplier or a colleague out of band from inside the thread, without the customer seeing any of it.
Pull specialists in from a chat tool. The handoff packet goes to your internal channel — a shared channel with the customer in it gets one line saying a specialist is joining.
Two threads that are the same problem become one. One thread that is two problems becomes two.
Targets with escalation actions, holiday calendars, and a countdown on every row so the queue sorts itself by what is closest to breaching.
Routing reads presence and current load, so work goes to somebody who can pick it up rather than somebody with a name.
On day one it is a setup path — the chat on your site, your knowledge, your email, your team — each step ticked off from what the platform can actually see, and a button to start a test chat. A test chat is a real conversation marked as a test, and it is left out of every figure in your reports.
On a quiet Tuesday it says so plainly, shows today’s numbers, and offers up to three things worth a look, each one linking to the screen where you would fix it.
Everything the AI is handling right now, before it reaches the queue.
Live boardThe actions those drafts are waiting on, as their own list.
ApprovalsThe SLA policies, business hours and routing behind the queue order.
WorkflowsHow chat, email, SMS, voice and RCS all land here.
Channels
Put the AI in front of every conversation, and keep your team in charge of everything that matters.