Skip to main content

Call control and voicemail

Module goal: set the call timers so the agent does not hang on the line for hours, does not talk to voicemail, and introduces itself properly to the callee’s call screener.
The call timers are on the Settings tab of the editor’s left panel, in the Call card (recording and call control). In detail — Recording and call control.
The Call card: recording, the duration cap, ringing, a silent caller, pause on request, answering machine detection and call screen handling

The Call card: the duration cap, ringing, a silent caller, pause on client request, answering machine detection and call screen handling. The timers in one card are the agent's manners on the line.

The Call card: recording, the duration cap, ringing, a silent caller, pause on request, answering machine detection and call screen handling

Call timers

The order is: reminders first, then the call ends when the silence threshold or the reminder limit is reached — whichever comes first; the duration cap applies independently of everything else.
Noise on the line does not suspend these timers for long. While the detector hears sound, the platform assumes the caller may be speaking and holds the silence count. But if recognition returns no words meanwhile, there is no point waiting forever: the hold is capped — about 5 seconds with recognition that returns words as the person speaks, and 30 seconds with recognition that returns only the finished result. After that the timers run as usual.
There is also a platform ceiling on top. Even if you chose No limit, a session longer than the platform ceiling is force-closed together with its telephony room, even if participants are still on the line. The default is 4 hours. The ceiling is a safeguard against stuck calls: without it a session could hang in the history for dozens of hours and skew reports. An unanswered call has an empty duration.

Answering machine detection (AMD)

The AMD parameters window: the Where to detect toggle, the What to check group, the detection timeout and the agent voicemail phrases

Answering machine detection (AMD). The agent greets at once, the detector listens in the background and, once it recognizes a voicemail, cuts the agent off — so a fragment of the greeting may end up in the mailbox.

The AMD parameters window: the Where to detect toggle, the What to check group, the detection timeout and the agent voicemail phrases
Detection is turned on with the Detect answering machines switch in the Call card; it is off for a new agent. It works on outbound phone calls; inbound calls are checked only with the Detect on inbound calls too toggle on in AMD parameters — it is meant for a campaign your company’s PBX dials itself, not for a line customers call on their own. Test and browser calls from the cabinet are not checked. The signs of a voicemail are chosen in the What to check group of the AMD parameters window. Phrases always work: the organization’s list on the Voicemail phrases page plus the agent’s own phrases. The organization’s recommended list already has typical voicemail greetings in Russian, Kazakh, English, Spanish, Uzbek and Kyrgyz, so it fits calls both in Kazakhstan and in the USA. The rest is turned on with checkboxes: the model on the first recognized words, the provider signal, the monologue — long speech without pauses — and silence on the line, off by default. There are three actions — Hang up (the default), Leave message (the agent can wait for the voicemail greeting to end) or Continue the conversation as is. In detail — on the Answering machine detection page.
Pitfall: AMD is wrong both waysDetection that is too aggressive cuts off live people who answer slowly (“hello?… yes?”, «алло… да? слушаю»); detection that is too soft leaves the agent chatting with a voicemail. This is the one setting that you must check with a live phone call to a real number before launching a campaign: to your own mobile, twice — once answering and once letting it go to voicemail. A test call from the browser will not help here: detection does not run on it.

The voicemail message

With the Leave message action, the text goes into the Agent voicemail message field: static text with {{var}} variables, or a prompt the model uses to write the line. An empty field means the built-in message in the agent’s primary language: it exists in Russian, Kazakh, English, Spanish, Uzbek and Kyrgyz. The Wait for the greeting/beep to end before speaking switch is on by default. An example of static text for an English-speaking agent:
Hi , this is Anna from Bright Dental, calling about your appointment on . Please call us back at .
And for a Russian-speaking agent:
Здравствуйте, ! Это Анна из клиники «Здоровье», звоню насчёт вашей записи на . Перезвоните нам, пожалуйста, по номеру .
{{first_name}}, {{visit_date}} and {{callback_number}} here are call variables you pass yourself, for example from the campaign’s contact data (see the Variables lesson). Whether a message may be left on a given call is decided by the organization’s voicemail message policy; a campaign can set its own on the Dialing step. The modes are Every voicemail reached, By rules and Never. If the policy blocks the message, the voicemail is still detected, the call ends without a message with the outcome Voicemail, no message left, and the attempt counts against all limits. The policy applies only to campaign calls. For organizations on the US market profile (us_tcpa), legal caps on the number of attempts apply on top of your policy, and the strictest limit wins.
Tip. The recommended policy setting is at most one voicemail per contact per day. In the message itself, name the company and the callback number, and leave out details a stranger should not hear: the mailbox may be played by someone other than the person you called.

Call screeners

Many phones answer an unknown number with a call screener — the callee’s assistant: iOS Call Screening, Call Assist, Cloaked AI. The assistant asks who is calling and why, then connects the call, holds the line, declines it or offers to leave a message. To the answering machine detector such an assistant is neither a person nor a voicemail. The Call screen handling row of the Call card — right under Detect answering machines — teaches the agent to tell a screener and a key-press menu from a person and a voicemail. It is off by default and works only on outbound phone calls, independently of answering machine detection.
  • A screener is recognized by phrases from the organization’s Call screener phrases list. Until the organization saves its own list, the recommended list of its market applies: for the US market — 20 phrases in English and 6 in Spanish, for other markets — 6 phrases in Spanish.
  • The agent answers “who is calling” with a template, the First reply (template), built from the facts of the Agent identity group in the Call screen handling window: Agent name, Company, Purpose of the call and Callback number. Keep the purpose of the call general: the owner of the phone sees the transcript.
  • The screener connects a person — the agent starts the flow from the first step in the same call. Declines the call — the outcome is Declined by a call screener. Offers to leave a message — the agent does the same as for a voicemail.
  • The mailbox is full (“The mailbox is full” and similar built-in phrases) — with call screen handling on, the agent does not speak the message, and the call ends as a voicemail.
In detail — in the Call screeners section.
Tip. For calls in the USA, turn Call screen handling on: screeners are common on US numbers, and without the handling the agent takes a screener for a person or a voicemail. For other markets, turn it on if the call history shows assistant lines. Before launch, add one more check call — to a phone with unknown-caller screening turned on.