AI Outbound Calling with Real Campaign Controls
Outbound calling fails on consistency and on compliance, usually in that order. Zemora runs outbound as structured campaigns: each one carries its own script, qualifying criteria, calling hours, attempt limits and concurrency cap, and every attempt ends with a recorded outcome and a reason.
- Per-campaign scripts, criteria and goals
- Attempt limits, retry spacing and calling hours enforced
- Concurrency capped per campaign
- Consent recorded; Do Not Call checks before dialling
Volume is not the hard part
Any dialler can make a lot of calls. What degrades is everything around the call: the script drifts, the fourth qualifying question gets skipped when the rep is tired, half the list never gets a second attempt, and the reason a contact was dropped is never written down – so the same list gets reworked from scratch six months later.
Compliance degrades the same way. Calling windows slip, Do Not Call checks get treated as a quarterly job rather than a pre-dial one, and consent lives in someone’s memory instead of a record.
What you actually configure
These are fields on the campaign, not settings buried in an agent persona.
Script and goals
Each campaign holds its own script and its own definition of a successful outcome. Two offers means two campaigns, not one agent trying to hold both.
Targeting criteria
The qualifying criteria that decide whether a contact is a fit, evaluated during the conversation rather than in a review afterwards.
Calling hours
Permitted windows per campaign, applied in the contact time zone – the control that keeps a national list from being dialled at breakfast.
Attempts and retries
How many attempts a contact gets and how far apart, so a list is worked properly instead of being redialled at random.
Concurrency cap
A ceiling on simultaneous calls per campaign, so a burst never outruns the humans available to take transfers.
Queue and scheduling
Contacts sit in a campaign queue with attempt counts, scheduled times and last-attempt timestamps, which is what makes the list state auditable.
What happens once someone answers
The agent opens with the campaign greeting, identifies itself as an AI agent, and works the discovery questions in a conversation rather than a fixed reading order – handling interruptions, tangents and the objections you told it to expect.
Qualified contacts are booked or transferred. Everything else is closed out with a structured outcome and a reason, and the whole exchange is stored as a transcript with speaker labels, so you can check what the agent said rather than trusting a summary.
- Discloses that it is an AI agent - never presented as a person
- Sub-400ms median response latency on live calls
- Transcript stored turn by turn with speaker attribution
- Structured outcome and reason recorded on every attempt
- Warm transfer to a named person when the call warrants it
The rules this operates inside
Automated outbound is regulated on both sides of the border and the rules are not optional. For voice we follow CRTC rules in Canada and the FCC Telemarketing Sales Rule in the US, including required disclosures and Do Not Call list checks before dialling. Consent is stored as a record against the contact, not assumed from a list purchase.
For text, Zemora is registered for A2P 10DLC with TCPA-compliant consent flows, automatic STOP and HELP handling, and per-campaign content review. Encryption, retention and access control are set out on the security overview.
We will not help configure a campaign that calls people who have not consented to be called, and calling-hour and attempt limits exist to be used rather than maximised.
Humans stay in the loop
Campaigns start in draft. Pilots run on a small segment with your supervision before volume opens, and every call is reviewable afterwards through its transcript and outcome. Sentiment scoring flags the calls worth listening to rather than asking anyone to sample at random.
Where the conversation should not be handled by an agent at all, transfer to a person is a capability set per agent – and it defaults to a warm handover with the reason recorded.
Where outbound fits
Outbound calling is one mode of Zemora AI Voice. The same agents answer inbound calls, and the qualification logic they apply is covered on the lead qualification page. Where the goal of the call is a meeting, see appointment setting.
For contacts who will not pick up a phone, AI SMS works the same list on the same contact record.
Outbound calling questions
Can I use different scripts for different campaigns?
Yes - the script, the goals and the targeting criteria are all held on the campaign itself, so campaigns running at the same time stay fully independent.
How many calls can run at once?
Each campaign carries a maximum concurrent calls setting. It is a deliberate ceiling: outbound volume that outruns the humans available to take transfers produces a worse result, not a better one.
Does it respect calling hours and time zones?
Yes. Calling hours are configured per campaign and applied in the contact time zone.
How are repeat attempts handled?
Maximum attempts and retry interval are campaign settings. Contacts sit in a queue with an attempt count and a scheduled next attempt, so the list state is auditable rather than guessed at.
Does the agent say it is an AI?
Yes. It discloses that it is an AI agent and is never presented as a human. That is a requirement under the disclosure rules we work inside, and it is also how the product is designed.
What about Do Not Call lists and consent?
Do Not Call checks run before dialling, and consent is stored as a record against the contact. For US SMS we are registered for A2P 10DLC with TCPA-compliant consent flows.
Can it transfer a live call to a rep?
Yes, where that capability is enabled for the agent. Transfers default to warm and are recorded with a reason and a timestamp.
Scope an outbound pilot
Bring a list, an offer and the criteria that make a contact worth a conversation. We will scope a campaign and start on a segment, not the whole file.