CEDX Live Chat · Customer Service
Live Chat treats chat as a floor, not a widget: every conversation carries a wait clock, every queue carries staffed-versus-goal, and the forecast engine warns about the wait you are about to have — not the one you already caused.
The best alert on the overview is a prediction: the Success queue's projected p50 wait will pass its goal under current arrivals — risk 63%. That is a staffing decision delivered while it is still cheap.
Runs on demo data. Visitor and company names in the captures are the demo seed's, not customers.
What it is
The chats list sorts the pain first: C-1312, a webhook-retries question in the mobile-app queue, has waited 8.7 minutes at urgent. Tabs split the floor into active 207, waiting 42, unassigned 60, missed 30 — four honest counts.
Each queue shows depth, p50 and p90 wait, staffed count and goal in one row. Support runs p50 2.4m against a 2.0m goal with 8 staffed; Technical is at 3.1m against 3.0m with 3. Understaffed is a filter, not a post-mortem.
One hundred eighty ratings with the distribution drawn: 8 one-star, 16 two-star, then the climb to 62 four-star and 54 five. The 154 comments come verbatim — “Fixed SSO same day” — with agent, queue and topic attached.
Product tour
Not a mockup and not a concept deck. This is what opens at /app/live-chat.
01 — Overview
Floor health is occupancy — 59% — with live active 101, waiting 42, unassigned 60 and 10 overloaded agents behind the ring. The forecast engine projects wait risk per queue: Success at 2 minutes with 63% risk, before it happens.
02 — Chats
Three hundred twenty chats in the workspace with topic, visitor, company, status, priority, queue and wait per row. The top of the waiting list is the argument for the next assignment, not the loudest channel.
03 — Queues
Twenty-four routing groups with depth, p50, p90, staffed and goal. Hardware answers in 6 seconds with 6 staffed; the VIP queue holds 54 seconds with a goal of 3.5m — the table shows which queues buy experience with headcount.
04 — CSAT
Seventy-five per cent CSAT on 180 rated chats, with the 24 one-and-two-star ratings filtered into their own review view. The comments name specifics — seat adds, SSO fixes, bot handoffs — which is what makes them coachable.
Who runs it
Roles, not references. We have no named customers yet, so nobody in these photographs is quoted, credited or claimed as one.
Owns the waiting column — 42 now, 7 past five minutes — and the staffing call that the forecast keeps forcing.
waiting · 42Watches containment by flow. The checkout bot at 22% against 40% overall is theirs to fix, and the alert names the hour it spiked.
containment · 40%Reads the 24 low-rated chats with the comments open, then rewrites the macro or the routing rule that caused them.
1–2★ · 24The shape of it
Every figure below is legible in the captures above. Nothing here is a projection of your estate — it is the state of the demo data.
How it runs
Chats land on queues, not agents — 24 routing groups, each with a staffed count and a wait goal.
The engine projects p50 wait against goal under current arrivals; Success is flagged at 63% risk before the wait exists.
Bots contain 40% of weighted flows, and the exceptions are named — the checkout bot's handoff spike is an alert, not a surprise.
CSAT closes the loop: 180 ratings, comments attached, and the 24 low ones filtered into a view of their own.
One record
Same demo floor, same workspace, same Ask AI button — a conversation that outlives chat has somewhere to go.
A chat that outlives its queue becomes a ticket. Desk's 201-ticket queue — same Northline demo estate — is where long threads land.
CEDX Knowledge BaseContainment is only as good as the library behind the bot. The 40% the bot contains are answers from those articles.
CEDX VoiceWhen the chat becomes a call, Voice is the floor it moves to — 88 calls live there in the same demo window.
CEDX CopilotAsk AI — ⌘J — sits on every chat screen. The Copilot console versions and traces the prompts behind it.
Limits
Finding this out on the third call is worse for you than reading it here, and worse for us.
Start
Open it right now
Your widget, your queues
Chat with the rest of service
Questions
Yes. Every screenshot is a capture of the running build, open at /app/live-chat. It runs on demo data — the visitor and company names in the rows are the demo seed’s.
The forecast engine projects each queue's p50 wait against its goal under current arrivals and prints a risk percentage — the Success queue reads 2 minutes at 63% risk on the overview. It warns before the breach, which is the only warning worth having.
An abandoned wait: 30 in the last 24 hours in the demo, counted on their own card and in their own tab on the chats screen — not folded into a green average.
It is printed per flow, and the failures are flagged. The overview carries an alert that the checkout bot's containment fell to 22% during a peak, against 40% overall — the spike is named, not averaged away.
No certification has been issued. What we can evidence about hosting, encryption and access is on the security page.
Live build, demo data, no card. Then ask which of your queues would be over goal right now.