Skip to main content

Call flow steps reference

Steps come in four groups. Ask the caller steps speak and listen. Decide steps branch without the caller noticing. Do steps run something in the background. Send the call steps end the flow by handing the caller somewhere.

Every step has a Step name shown on the canvas. Steps that speak accept either What to say, read out in your carrier's synthesised voice, or a recording of your own under Or play a recording instead.

Recordings on a speaking step​

Your carrier's synthesised voice is one voice and it cannot be changed per step. Any step that speaks can play a recording in its place, so the flow can sound like your business: your own voice, your own accent, your own language, or a jingle.

Click the field under the step's What to say box and pick one of four tiles.

TileWhat it does
RecordRecords from your computer's microphone, up to 2 minutes. A level meter shows the microphone is live, and you hear the result before you keep it.
UploadTakes an MP3 or WAV you already have, up to 10 MB. Anything your browser can decode is converted for the phone line.
Use a voiceTypes a line and has it read by one of the voices from an AI engine you have connected.
Use a linkPlays a public MP3 or WAV from a URL of your own. Nothing is stored here, so keeping that file online is up to you.

Once a step has a recording, the field becomes a small player. Replace swaps it for another and Remove goes back to reading the text.

Leave What to say filled in even when a step plays a recording. It is what callers hear if the file cannot be fetched at call time, which is the difference between a caller hearing the wrong greeting and hearing nothing at all.

Use a voice​

Reading a line with an AI engine uses the same engines, the same voices and the same API key as your AI voice agents, so a menu can sound like the AI agent sitting behind it.

  1. Pick the AI engine
    • Shown only when you have connected more than one. With a single engine connected there is nothing to choose and the line under the box names it.
    • The engines do not sound alike and they do not cover the same accents. Sarvam AI is the one with Indian voices, ElevenLabs Agents has the largest library, and OpenAI and xAI Grok each offer their own set. Pick per prompt, not per flow: nothing stops one flow using two.
    • Your choice is remembered, so the next prompt you write starts on the same engine.
  2. Pick the voice
    • The list is the voices on your own key. Press the play button beside it to hear one before spending a generation on the real line.
  3. Check the text
    • It starts as whatever is in What to say, so most prompts need no typing. Up to 1500 characters.
  4. Click Generate
    • The line is read, saved, and put straight into the step's player.
What this costs

Voices and generations are billed by the AI vendor on your own key, at their rates. Nothing is added by WhautoVoice, and none of it appears on a WhautoVoice invoice. Generating a prompt is a few seconds of audio written once, not something that repeats on every call.

If you have no AI engine connected, the tile explains that and links to the page where you connect one. Record, Upload and Use a link work without one. On ElevenLabs the key also needs the Text to Speech permission: a key without it lists your voices perfectly well and still refuses to read a line. See Connect an AI engine.

Switching engine clears the selected voice. Voice names belong to the vendor that issued them, so an ElevenLabs voice cannot be carried across to Sarvam.

Which steps take a recording​

StepField
MenuOr play a recording instead, and Or play a recording after an invalid key
Collect digitsOr play a recording instead, and Or play a recording after an invalid entry
SayOr play a recording instead
AI voice agent, Call queue, Ring an agent, Forward to a numberOr play a recording while connecting
VoicemailOr play a recording before the beep
Hang upOr play a recording as the goodbye

Ask the caller​

Say a prompt and branch on the key the caller presses.

FieldDefaultNotes
What to sayrequiredMention every key you list below.
Seconds to wait for a key51 to 30.
Retries before giving up20 to 5. After the last retry the No input / invalid path is taken.
Said after an invalid keyoptionalPlayed before the prompt repeats.
Times a caller may come back to this menu51 to 20. Protects against a repeat key looping forever.
Keys the caller can pressat least oneDigits 0 to 9, star and hash. Each key has a label and a destination. Add a repeat key adds a key that replays this menu.

Paths: one per key, plus No input / invalid.

Collect digits​

Capture a number (account id, PIN, date of birth, callback number) into a variable that later steps can read.

FieldDefaultNotes
What to sayrequiredTell the caller what to enter and which key finishes.
Save the digits asrequiredThe variable name, for example callbackNumber.
Minimum digits1
Maximum digits20
Finish keyhashThe key that ends entry early.
Seconds to wait5
Retries before giving up2
Said after an invalid entryoptional

Paths: Collected (enough digits were entered) and Gave up (the retries ran out).

Say​

Speak a message, then continue.

FieldDefaultNotes
What to sayrequiredOr a recording.

Path: Then.

Decide​

Business hours​

Take a different path outside opening hours.

FieldDefaultNotes
SchedulerequiredOne of your schedules from Settings > Business hours. Holidays count as closed.

Paths: Open and Closed. See Business hours.

Condition​

Branch on the caller, their tags, or a value collected earlier.

FieldDefaultNotes
Conditionsat least oneEach row is Look at a field (the caller's tag, stage, a custom attribute, or a variable from Collect digits), an operator (is, is not, is one of, is not one of, contains, does not contain, greater than, less than) and a Value.

Paths: Matches and Does not match. Use it to send VIP callers straight to a person, or to route a returning patient differently from an unknown number.

Do​

Run an automation action​

Tag the contact, notify the team, call a webhook, change the pipeline stage, assign an owner or schedule a callback. These are the same actions the automation rules use. The action runs in the background and the call continues immediately.

FieldDefaultNotes
ActionrequiredAdd tag, Remove tag, Update stage, Update notes, Send webhook, Notify team, Schedule callback, Assign contact owner.
Parametersdepends on the actionFor example the tag name, the webhook URL and headers, or the callback delay in minutes.

Path: Then.

Send the call​

These steps end the flow.

AI voice agent​

Hand the caller to an AI agent.

FieldDefaultNotes
AI agentrequiredOne of your AI voice agents.
Said while connectingoptionalOr a recording.

The phone number must be registered with the AI agent's AI engine where that engine needs it. See Connect an AI engine.

Call queue​

Put the caller in a queue for the next free agent.

FieldDefaultNotes
QueuerequiredCallers hear hold music until an Available or Wrap-up member picks up.
Said while connectingoptional

Ring an agent​

Ring one specific team member.

FieldDefaultNotes
AgentrequiredIf they do not answer, the caller reaches voicemail.
Said while connectingoptional

Voicemail​

Take a message.

FieldDefaultNotes
Prompt before the beepoptionalLeave blank for the default greeting.

The message appears as a voicemail on the call, and a New voicemail notification goes to the workspace.

Forward to a number​

Forward the call to an external phone number, for example a mobile or another office.

FieldDefaultNotes
Phone numberrequiredInternational format, for example +1 202 555 0100.
Ring for (seconds)30
Said while connectingoptional

Hang up​

Say an optional goodbye and end the call.

FieldDefaultNotes
Goodbye messageoptional

Validation messages​

When you click Activate, the builder lists everything that would stop the flow working. Each item names the step it belongs to.

MessageWhat to do
The flow has no steps yet.Drag at least one step from the palette.
No first step is set - pick which step answers the call.Right-click a step and choose Make this the first step.
(field) is required.Open the step and fill in the named field.
add at least one key the caller can press.The menu has no options. Click Add key.
the key (n) is used twice.Two options share a key. Change one of them.
"(path)" points at a step that no longer exists.The destination was deleted. Choose a new destination or remove the path.
this step does not lead anywhere yet.A path is not connected. Every path must end in a Send the call step or another step.
"(path)" points back at this step, which would loop forever.Change the destination. Use Add a repeat key if you want a menu to repeat.
every key repeats this menu - give the caller somewhere to go.At least one key must lead out of the menu.
Need help?

Email us at hello@whautovoice.com and we will get you sorted.