The Voice channel

Voice turns Atender into a cloud call center. Customers dial a number you provision, an IVR flow you design routes them, and the call lands in a queue staffed by human agents — or hands off to an AI voice agent — all while the conversation lives in the same inbox as Email, Web Chat, and every other channel.

What you get

• Phone numbers in the markets you serve. Request a number per country from inside Atender. Twilio handles provisioning and regulatory compliance behind the scenes.
• Visual IVR builder. Drag-and-drop nodes onto a canvas to design every call flow — greetings, menus, conditions, transfers, voicemail. Every node type you can add maps to a real call-handling action. A recording node from an older flow does nothing; recording is a voice setting.
• Call queues with full queue experience. Each queue is staffed by a team you assign and can give agents a wrap-up time between calls. Hold audio and comfort announcements come from the global queue settings, and a queue plays its own when you turn on Override global settings for it. A queue can also carry a Max hold limit — the longest a caller waits — and, when the wait ends, offer a call back or move the caller to another queue, so queues chain into an escalation ladder. See Set up queue overflow with Max hold.
• Callback. Instead of forcing callers to wait on hold, offer them a callback — they hang up, and the callback waits in the queue. By default a callback keeps its place in order of arrival. Call queue order, under Global on the Queues tab, can put callbacks behind every call or ahead of every call in the same call queue instead. A queue that prioritizes calls by customer tier still serves higher tiers first. When an agent is free, the agent gets a callback popup and clicks Call Back, and Atender dials the customer.
• AI voice agents. An IVR node can hand the call off to an Agent Stack that runs the conversation autonomously with a real-time LLM loop. When the AI decides to escalate, it transfers the call to a human queue, but only if the Agent Stack has a queue or a team for voice handover. When the AI cannot continue, or the transfer fails, the stack’s fallback runs: Send caller to a human queue, Take a voicemail, or Hang up.
• TTS in 33 locales (24 distinct languages). Type your message, pick a voice and language, and the IVR speaks it. Upload your own audio if you prefer a professionally recorded greeting.
• Accent translation. Make an agent’s accent softer, in English, for the caller on calls that an agent answers in the browser. An admin turns it on for a call queue, and each agent turns it on for themselves. A caller whose language is known and is not English is never translated. It replaces AI noise suppression on that agent’s calls and adds a delay of less than one second. See Turn on accent translation.
• Call recording and transcription. Configure recording for the whole tenant, and override it on a Send to Queue step. Callback and outbound calls are always recorded, whatever the tenant setting; on an outbound call, the customer can still press 9 at the recording notice to opt out. Consent and opt-out are built in. Atender can transcribe recorded calls.
• Opening hours awareness. An IVR node branches the flow based on your business hours — open, route to the queue; closed, send to voicemail or an alternate path.
• Queue announcements while callers wait. A queue can play comfort announcements — the first after a delay you set, then one every 90 seconds — and they can include the caller’s queue position and estimated wait. Each announcement is enabled individually on the Queues tab — a disabled one is never spoken. An announcement speaks in its own TTS language and voice if you set one, otherwise the call’s. Because an announcement repeats while the caller waits, Atender speaks it with a Standard voice: a premium voice is replaced by its Standard version, or by another Standard voice for that language, and a language with no Standard voice keeps the voice it has; a language outside the voice catalog is spoken with the source text instead of a translation. When you edit the text, {{estimatedWaitTime}} expands to a whole phrase such as “in 2 minutes”, so do not write a time word before or after it — some languages put that word at the end. That phrase is normally spoken only when Atender can measure the queue wait from live queue state or queue history; if no measured estimate is available, an announcement containing {{estimatedWaitTime}} is skipped, while a position-only announcement can still play. The exception is the older hold path — used, for example, when an AI voice agent that did not request a transfer falls back to a queue — where Twilio reports the caller’s position itself: if nothing has measured the wait there, the phrase falls back to a flat “in 2 minutes”.

How it fits

Every voice call lands in Conversations like any other channel. It carries the same lifecycle (Active, Snoozed, Done, Archived), can be assigned, tagged, and merged, and it surfaces in the same inbox views your team already uses. When a caller hangs up after the IVR sends the call to a queue or to the AI voice agent, and before anyone answers, the call is a missed call. A call the AI voice agent answered is not a missed call, even if the caller hangs up before a transfer. By default the conversation closes to Done and shows a Missed badge. Turn on Keep missed calls open in Global voice settings to keep a missed call in Active with its Missed badge instead, so your team can call the customer back. Agents see live call controls — mute, transfer (on a normal transferable live call), a dialpad, and end call — while the call is active, and the conversation stays after the call ends, with the recording and transcript when the call was recorded. A call that goes through a Send to Queue node follows the tenant recording default, which is off, and the node can force recording on or off. Callback calls and outbound calls are always recorded. Callback calls still show the standard in-call controls, but the Transfer button is absent because callbacks cannot be transferred from that call state.

A voice call can also start as a callback from a different channel — a visitor requests a callback from a form or the public callback widget, and a new voice conversation is created automatically.

What is NOT in the Voice channel

A few things worth being explicit about so you don’t go looking for them:

• No native dialer for outbound campaigns. Voice is built around inbound calling and callbacks. An agent can call one contact at a time from CRM, but there’s no power-dialer or list-import for outbound sales calling.
• Recording is not configurable per call. Recording is decided by the tenant default or the Send to Queue node — not toggled mid-call by the agent. Callback and outbound calls are always recorded. If one queue records calls differently from your settings, contact Atender support.

Where to start

If you’re setting up voice for the first time, follow this order:

1. Add a phone number — pick a country, submit compliance details, and wait for the number to be provisioned.
2. Build an IVR flow — design the call experience on the visual canvas.
3. Set up a call queue — pick a team, configure timeouts, and assign the queue to your IVR’s Send to Queue node.
4. Configure TTS voices — pick the language and voice the IVR speaks in.

After that, dial the number from your phone and walk through your own IVR — it’s the fastest way to find rough edges before customers do.

See also

• IVR node types
• Build an IVR flow
• Set up a call queue
• Configure callback