How it works · Outgoing Voice

Rules first. Then the call.

We never just “hit dial.” Every attempt goes through your campaign settings and a safety check. If something fails, nothing rings — and we log why.

Set up the call type

Create a campaign. Choose the purpose, mobile or landline, whether permission is required, the opening lines, and what the AI must not say.

Use people you’re allowed to call

You bring the relationship—someone who booked, joined, or opted in. Numbers that opted out go on your block list. We do not build a cold-list product for you.

We check before dialing

Is the campaign on? Is the number blocked? Do you need permission, and do you have it? Have you called too often? If anything fails, we stop and record the reason.

The agent runs a controlled conversation

If the check passes, the AI voice call runs. It introduces itself (who you are + that it’s AI + why it’s calling), does the job (for example, confirm an appointment), respects “stop,” and can request a human callback.

Your team finishes the loop

Staff work the callback list, review logs and transcripts, and improve scripts—without losing the safety rails.

Example

A Tuesday appointment confirm

One realistic path from list to disposition — what “good” looks like without drowning in product jargon.

1
List + tags

Jane is on your outbound list with consent tagged for appointment reminders (you own that relationship and the lawful basis).
2
Pre-dial gate

Campaign is enabled, number is not suppressed, destination class is allowed, consent artifact is present, frequency caps are fine → allow. If any check failed, no carrier dial — attempt is logged as blocked.
3
Forced open

She answers. The agent speaks identity, that the voice is AI, and the purpose — before free-form chat.
4
Three ways it can end well

“Yes, that works” → outcome confirmed. “I need a person” → callback ticket for your staff (MVP is callback, not a live warm transfer). “Stop calling me” → opt-out and suppression so the next attempt never passes the gate.
5
Review

Your team sees the call on Call review: times, disposition, and (when live) transcript and optional recording under retention policy.

On the line

The “AI agent” is not a free chatbot with a phone. It is a short loop bounded by your campaign and platform rules.

Listen

Speech is turned into text so the agent can understand the person — and so you can review a transcript later.

Decide

Policy and campaign intent guide the reply: allowed topics, blocked topics, and required tools (opt-out, escalate, end).

Speak

A brand AI voice answers — never a clone of a real person, celebrity, or candidate. Opening disclosure is forced.

Tools

Confirm-style outcomes, do-not-call, human callback ticket, hang up cleanly. Later: domain tools you allow (e.g. lookup appointment).

Scripts and prompts are not one marketing paragraph. Your campaign sets the opening and what may be discussed; the platform keeps hard rules (AI disclosure, opt-out, no impersonation). How those layers change agent behavior is documented for operators inside the product under Agent guide.

What your team still does

Lists & permission — who to call and why you’re allowed to call them.
Callback SLA — when the agent escalates, a person returns the call on your schedule.
Campaign & script changes — purpose, open, blocked topics; improve from real reviews.
Edge review — escalations, opt-outs, and odd transcripts for coaching and counsel.

What we never do

Clone a real person’s voice or impersonate a celebrity or candidate.
Spray unconsented wireless lists as a product motion.
Claim that “AI chat” removes automated-voice rules.
Certify that your consent language is legally sufficient — you own that; we enforce the policy you tag.

How we think about rules →

Take the guided product tour

Five stops: this story → dashboard → sample call review → agent guide (prompts & guardrails) → test dial (gate only — no call placed).

Start tour → Dashboard Restart from here What’s included Start a pilot