The four steps
Creating one walks four decisions, and each changes what the next means.1
Voice — what the call says
Three sources:
- A recording. Something somebody made and has listened to.
- Write it, a voice reads it out. Text-to-speech, spoken once and cached.
-
Write it, an AI delivers it. An agent profile
rephrases it in its own manner and voice — it reads dates and account
numbers the way a person does rather than reciting them.
2
List — who to call
A segment, or a CSV upload. Either way you see
the list before you continue: a segment’s first rows and its total, or the
file parsed into a table.
3
Channels
How many calls run at once. For a broadcast this is the whole of its
pacing — there are no agents to wait for, so it opens to this ceiling.
4
Start
Now, or at a date and time. Plus the daily hours it may dial in — it never
calls outside them, whatever else is set.A broadcast can also run a workflow after the message, so somebody who
needs to respond can press a key. Everyone who hears the message enters it,
so leave it empty unless you mean that.
Personalising the message
A written message can carry fields from each contact’s own row:Answering machines
FireTone waits briefly for a voicemail beep before it starts talking, because a message played over a greeting is a message nobody hears. No beep is the ordinary case — a person picked up — so it does not wait long. The call sheet distinguishesmessage_played from message_left_on_machine.