Campaigns & Deployments

A Deployment is the moment your Finn stops being a draft and starts making real calls. It takes a Finn you've built, an Audience of contacts, a phone number, and a schedule — and turns them into a running campaign.

Before you launch. Always test on yourself first (the "Test on Call" button on the Finn page). Always set a spend cap on your first few deployments. Always check the calling-hour window — there's no faster way to lose customers than calling them at 3 AM in their local time. The deployment screen makes all of this easy; just don't skip it.

Pausing is free. If anything looks weird mid-campaign, hit Pause. Active calls finish naturally, no new calls go out, no extra cost. Resume when ready.

The mental model

Finn  +  Audience  +  Phone Number  +  Schedule  =  Deployment

Think of a deployment as a campaign instance. The same Finn can run multiple deployments simultaneously against different audiences or schedules — they're independent.

Where to find deployments

• Dynamic Deployments — the launcher. Create new deployments here.
• Scheduled Deployments — deployments waiting to start at a future time.
• Live Deployments — deployments currently running. Real-time call counts and a live call table.
• Deployment Analytics — completed (or in-progress) deployments with full metrics drilldown.

Dynamic Deployments — where you launch new campaigns.

Live Deployments — real-time call counts and active call table.

Deployment types

Standard

Default. The Finn calls contacts one at a time (or in small concurrent batches, depending on your plan).

Best for:

• Moderate-volume campaigns (up to a few thousand contacts).
• Inbound number routing.
• Use cases where you want gentle pacing.

Parallel

Enterprise. The Finn calls thousands of contacts simultaneously.

Best for:

• High-volume outbound — debt collection sweeps, mass notifications, election outreach.
• Time-sensitive blasts where the entire audience must be reached in a tight window.

Parallel requires sufficient phone-number capacity (you can't dial 5,000 calls from one number). See Phone Numbers → for number pools.

Scheduled

Either type can be scheduled. The deployment sits in Scheduled Deployments until its start time, then automatically transitions to Live.

You can also configure calling-hour windows — even if the deployment runs for days, the platform only dials contacts during their local 9 AM – 6 PM (or whatever window you set). Contacts in different time zones get called at different absolute times.

Launching a deployment

1. Open Dynamic Deployments (or click Deploy from a Finn's page).
2. Configure each field below, then click Launch.

Required fields

Schedule fields

Advanced

Pre-launch checklist

Before hitting Launch on anything that costs real money:

• Test the Finn on yourself first (Test on Call).
• Audience has been validated (no invalid numbers, no test numbers from a prior import).
• From number is registered and SHAKEN/STIR attested in your country.
• Calling-hour window respects local-time-of-day for the contacts.
• Post-call analysis fields are defined — otherwise you can't measure success.
• You've reviewed the cost estimate the platform shows in the launch confirmation.

Monitoring a live deployment

Live Deployments shows:

• Active calls — currently on the phone.
• Completed calls — finished, with status (success / failed / no-answer / voicemail).
• Success rate — completion-weighted moving average.
• Buffer status — whether the dispatcher has runway. A draining buffer means you might run out of contacts.
• Per-call table — live-updating list of every call. Click to open the Data Extractor Sidebar with transcript + audio.

Health indicators (top of the page) show Redis buffer health, dispatcher status, and any queue backups.

Pausing & stopping

• Pause — stops dialing new numbers. Active calls finish naturally. Resume any time.
• Stop — same as pause, plus marks the deployment as completed. Cannot be resumed.

Active calls are never killed mid-conversation — the caller's experience is always preserved.

Retries & no-answer behavior

If a call fails (no answer, busy, voicemail-end), the platform schedules a retry based on your retry policy.

• Retries happen at the next valid calling-hour window for that contact.
• Retries respect the global retry cap (default 2).
• If voicemail behavior is Leave a message, the platform plays a recorded or TTS-generated message and does not retry.

Cost & billing

Each call incurs:

• Telephony cost — per-minute charge for the carrier leg. Varies by country.
• AI cost — per-minute charge for the LLM + TTS + STT. Flat by tier.
• Optional add-ons — premium voices, post-call analysis, recording storage.

The deployment confirmation modal shows estimated cost based on average call duration. Actual cost is computed per-call and rolled up nightly.

See Billing & Plans →.

Common pitfalls

Next

• Analytics → — measure how it went.
• Phone Numbers → — provision and pool numbers.
• Integrations → — write outcomes back to CRM.