CEDX Voice · Customer Service
Voice scores the things contact centres usually argue about: clarity per leg, drop risk per live call, who ended the media, and which queues are understaffed — as columns on the floor, not as a weekly report.
The hangup-side panel is the honest one: the network ended 81 calls in the window; callers ended 47. When your own infrastructure tops the blame chart, the chart is probably telling the truth.
Runs on demo data. The phone numbers in the captures are demo lines, not real ones.
What it is
The live floor shows every call by state — 32 talking, 16 ringing, 22 on hold — with age, risk and talk percentage. CL-1042 has been talking 4:22 at risk 100 with the agent at 78% talk time; the monologue alert named it before any supervisor would.
Direction, status, queue, agent, duration, clarity and risk on every row. CL-1033 failed on Technical at 34 seconds with clarity 2.1 — a network event you can diagnose, not a complaint you cannot. AI receptionists appear in the agent column, labelled as AI.
Depth, ASA, goal, abandon rate, staff actual-versus-needed and risk average per queue. Field support runs ASA 41s against a 20s goal; Partner desk abandons 8.3% at five-of-seven staffed.
Product tour
Not a mockup and not a concept deck. This is what opens at /app/voice.
01 — Overview
Offered-versus-answered across 30 days with events as markers — a new VIP number live on Jul 12, an AI receptionist version on Jul 21, a promo-week spike on Jul 30, the clarity floor raised to 3.5 on Aug 2. The hangup-side panel says who ended the media: network 81, system 58, caller 47, agent 47.
02 — Live floor
Eighty-eight live calls with state, queue, agent, age, risk and talk percentage. The wrap column lists who is in after-call work, and the wait-age bars split ringing plus hold into under-30s through over-2m.
03 — Calls
Three hundred twenty calls with caller, direction, status, queue, agent, duration, clarity and risk. The failed rows say why in the columns — 21 seconds, clarity 4.3, an AI agent — so triage starts from the row itself.
04 — Queues
Twenty-four queues with depth, ASA, goal, abandon, staff and average risk. Eight queues are over their ASA goal right now, 127 callers are waiting in total, and the understaffed toggle reduces the board to the 13 that need people.
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 floor health number — 80 today — and makes the staffing call when a queue's ASA passes its goal. The queue board in the photograph is the same job, older technology.
understaffed · 13 of 24Runs the numbers and the IVR flows. The menu-3 drop step and the packet-loss cluster on one line are theirs to fix.
menu-3 drop · 8.2%Works the high-risk list. A 78% agent monologue on a VIP call becomes a coaching note, not a blame.
high risk · 86The 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
Calls offer to queues through the IVR. The menu-3 step drops 8.2% of support calls before a queue — and the alert names the menu.
81.9% connect; the rest abandon or fail and stay counted, 42 rows in the dropped tab with causes in the columns.
Clarity is scored per leg — 3.9 on average — so “the line was bad” is a number with a floor, not a mood.
Eighty-six high-risk calls hold a live list; one VIP call's 78% agent talk time became the monologue alert a coach now owns.
One record
Voice shares its demo estate with the queue, the chat floor and the remote toolbox — same workspace, same Ask AI button.
Callbacks and complaints become tickets. Desk's queue — same Northline demo estate — is where the 42 dropped calls get worked properly.
CEDX Live ChatThe chat floor and the call floor trade overflow: 42 waiting in chat, 127 waiting on the lines, one staffing conversation.
CEDX RemoteVoice ends where screen-sharing starts. Remote runs those sessions with recordings and anomaly scores.
CEDX CopilotAsk AI — ⌘J — sits on every Voice screen, versioned and traced by the Copilot console 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 numbers, your queues
Voice with the rest of service
Questions
Yes. Every screenshot is a capture of the running build, open at /app/voice. It runs on demo data — the contact centre on the chrome is the demo tenant.
A per-leg quality score, 3.9 on average across the demo window. The floor was raised to 3.5 on Aug 2 — visible as a marker on the overview chart — and the packet-loss cluster alert names the number it affected.
Live and recent calls get a risk score; 86 sit at ≥55 and the worst read 100 on the live floor. The signals show up in the alerts: monologue risk from talk-time splits, network clusters from packet loss.
They are labelled in the agent column — billing, auth and a Spanish receptionist handle demo calls. They are demo configuration, shown because the floor shows them; we do not quote their containment as a figure for your centre.
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 who ended your last hundred calls.