Writing the call workflow
Turn a call into ordered stages with objectives, required slots, turn and time budgets, exit criteria and fallbacks — and know when supervision actually applies.
Most people write a voice agent's script as one long instruction and then wonder why the call wanders. The Workflow tab fixes that by making the call an ordered set of stages, each with a job to do, a budget to do it in, and a defined way out. It is the difference between a prompt and a script.
Before you start
You need an editable draft version of an agent and the voice edit permission. Published and archived versions are read-only — use Edit as new draft first.
Set the goal first
Before you add a single stage rule, open the agent's Agent tab and fill the two fields on its Prompt card.
Goal is the objective previews and test calls work toward — Campaigns use their own purpose; this goal drives previews and test calls. Approved next steps is the list of calls-to-action the agent is allowed to offer, one per line. The placeholder shows the shape:
Schedule a meeting
Request a callback
End politelyAnything not on that list is not on the table. That is the point of the field: it is a closed set, not a suggestion.
Add the stages
Every stage runs on its default instructions until a rule tunes it, so with no rules the Stages card reads No stage rules yet. A rule is how you tune one.
Click Add stage and pick a stage
The menu lists the stages without a rule. Picking one opens its sheet; the stage is fixed for that sheet.
Write the Objective
The placeholder reads Complete the <stage> stage. Replace it with one sentence about what must be true when this stage ends.
List the Required slots
Facts the agent must collect before this stage may advance — snake_case
keys, as the runtime stores them. Enter adds a slot; the placeholder is
need, timeline.
Set Max turns and Max seconds
The budget. Turns run 1–50 and seconds 5–1800; leave either empty to store no cap.
Write Exit criteria and Fallback
What signals the stage is done, and what to do when it cannot be done.
Click Add rule
The sheet's Tools row shows What the runtime hands the agent here. — the permitted tools this stage may use. It is read-only; the Tools tab decides which tools the agent may call at all.
Open a rule's row to change it and click Save rule. Remove stage in the sheet drops the rule after a confirm, and the stage goes back to its default instructions. The card's footer counts how many stages carry a rule.
The twelve stages
They run in this order, and the names describe what each is for:
| Stage | What it is for |
|---|---|
| Preflight | Before the conversation starts — no tools are handed over |
| Introduction and disclosure | Greeting, who is calling, and the AI disclosure |
| Permission to continue | Asking whether now is a good time |
| Discovery | Understanding the situation |
| Qualification | Establishing whether this is a fit |
| Value alignment | Connecting what they need to what you offer |
| Objection handling | Answering the pushback |
| Call to action | Offering one of the approved next steps |
| Booking or transfer | Actually booking, or handing to a person |
| Confirmation | Reading back what was agreed |
| Wrap-up | Closing the call cleanly |
| Post-call | After the conversation ends — no tools are handed over |
You do not need a rule on every stage. A callback-confirmation call might tune four. A cold discovery call might tune eight.
Settings reference
| Setting | What it changes | Default |
|---|---|---|
| Goal | On the Agent tab. The objective previews and test calls work toward; campaigns use their own purpose. | — |
| Approved next steps | On the Agent tab. The calls-to-action the agent may offer, one per line. A closed list. | — |
| Stage | Picked from Add stage; one rule per stage. | — |
| Objective | What this stage must accomplish. Up to 500 characters. | — |
| Required slots | Facts the agent must collect before this stage may advance. | — |
| Max turns | Turn budget for the stage. Range 1 to 50; empty stores no cap. | — |
| Max seconds | Time budget for the stage. Range 5 to 1800; empty stores no cap. | — |
| Exit criteria | What signals the stage is complete and the agent may move on. | — |
| Fallback | What to do when the stage cannot complete inside its budget. Placeholder: Move on politely without inventing information. | — |
| Tools | Read-only. The permitted tools the runtime hands the agent in this stage. | — |
Budgets are the part people get wrong
A stage's Max turns and Max seconds are a budget, not a target. Four turns is enough to ask a question, hear an answer, ask a follow-up and acknowledge it. If a discovery stage needs twelve turns, the objective is probably two objectives wearing one coat — split it.
The Fallback matters more than it looks. It is what the agent does when the budget runs out with a required slot still empty. The placeholder — Move on politely without inventing information — is the right instinct: the failure mode you are guarding against is an agent that invents the answer rather than admit it did not get one.
Write exit criteria as observable facts
“The person has said a rough timeframe” is testable. “The person is engaged” is not. The agent has to decide the stage is done using only what was said on the call.
When stage supervision actually applies
This is the part that surprises people. Writing stages always shapes the composed prompt, but deterministic stage supervision — the runtime actually enforcing the turn and time budgets, the listen-first timing, the interruption handling, the bounded memory and the silence policy — only happens when Workflow-supervised voice runtime is switched on in SettingsVoice & Telephony. The agent's Advanced tab repeats the workspace state at the top, and its Workflow supervisor switch under Runtime features can turn stage supervision off for one agent.
Off by default
Workflow-supervised voice runtime defaults to off, and the row only appears when the runtime settings are available in your environment. With it off, your stages still guide the agent through the prompt, but nothing forces a stage to end at four turns. Turn it on before you rely on the budgets, and it applies to new calls only.
Troubleshooting
| Symptom | Cause | Fix |
|---|---|---|
| Add stage is greyed out | Every stage already carries a rule, or the version is read-only | Open an existing rule to change it, or use Edit as new draft |
| The agent ignores a stage's turn budget | Workflow-supervised voice runtime is off for the workspace, or the agent's Workflow supervisor is off | Turn it on in Settings → Voice & Telephony and check Runtime features on Advanced; it applies to new calls |
| The agent invents an answer when it runs out of turns | Fallback was rewritten to push harder | Restore a fallback that moves on without inventing information |
| Validation fails against Workflow when you publish | A rule is missing a required field, or a budget is outside its range | The error names the field and the Workflow tab carries the badge — fix it and validate again |
| The agent offers something you never approved | It is not in Approved next steps, but the persona or objective implies it | Keep the offer in the approved list and out of free-text fields |