What call screening sounds like
The phone answers with a synthetic voice, typically saying the person is using a screening service and asking for your name and the reason for the call. The real person may join later, or never. The details and wording vary by phone, country and software version.
Why classic AMD gets it wrong
Timing-based AMD expects a person to say something short and wait. A screening assistant says a sentence and waits, which can look like a long greeting (MACHINE) or like a talkative person (HUMAN), depending on settings. Transcript-based AMD hears natural language that isn't a voicemail greeting. Either way, many screened calls reach agents.
How nomoVM handles it
nomoVM's model has a separate class for screening, learned from real screened calls. Screened calls come back as AMDCAUSE=SCREENING with AMDSTATUS=MACHINE, so by default they're handled like any machine. Because the cause is separate, you can choose a different route for them in your dialplan:
exten => 8370,n,GotoIf($["${AMDCAUSE}" = "SCREENING"]?screened)
; ...
exten => 8370,n(screened),Playback(custom/who-we-are) ; say who you are, then hang up or waitSome centers leave a short message that names the company, since the screened person often reads the transcript later. What's right for you depends on your campaign and the rules that apply to you.
What to watch in your reports
- The share of SCREENING among answered calls, per list and per time of day
- Whether screened numbers answer on a later attempt
- Agent feedback on "robot" calls that still get through
More on the other checks nomoVM runs: how AI answering machine detection works.