Zemora AI Voice - private beta

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.

The problem

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.

Campaign controls

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.

On the call

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.

Consent and compliance

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.

Oversight

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.

Common questions

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.