Everything teams ask before their first campaign.

About ten minutes. Create a workspace, paste your script into an agent, and place a test call to your own phone. Interrupt it, answer vaguely, switch languages — then read the transcript it produces. That output is the product, and it is the only thing worth judging.

For test calls, no. For a campaign, yes — rent an Indian +91 number from us in a couple of clicks, or connect a Plivo account you already own and keep your existing caller ID.

No. Campaigns run from a CSV upload and the dashboard. A developer is only needed if you want calls triggered from your own system — that path is documented too.

A system prompt describing what the agent should find out, and a welcome line. Both are ordinary English. The first version is never the last — expect to read fifty transcripts and rewrite it twice.

Prepaid and per call-minute. You top up a wallet, calls draw it down, and dialling stops when it is empty. No subscription and no invoice at month end. Full plan comparison.

Telephony, speech-to-text, the language model and the voice. There is no second bill from a provider to reconcile — unless you deliberately bring your own keys, in which case you pay those providers directly.

You are charged for connected time. A call that rings out has no conversation minutes to bill.

The rate follows your top-up. A larger balance moves you to the Growth or Enterprise tier and the per-minute rate drops for every call after that — there is nothing to sign.

It sounds like a clear, slightly formal person on a phone line. What gives an agent away is rarely the voice — it is a reply that arrives too late or ignores what was just said, which is why turn-taking gets more attention here than voice selection.

Yes, on two-way calls. The agent stops speaking and listens, as a person would.

English, Hindi and Telugu today, and the agent follows the language the caller uses rather than forcing one. Why telephone audio makes this harder than it sounds.

A knowledge base scopes what it may state as fact, and the prompt sets hard boundaries — no prices it was not given, no promises, no advice. Calls needing judgement are designed to capture the situation and hand over to a person.

Treat them as a consistent first-round screen, not a final decision. The same rubric grades every call, which is more than a tired reviewer manages at 6pm — but spot-check transcripts for your first campaigns and tighten the wording.

Yes, within rules on consent, calling hours and do-not-call lists. Calling windows and opt-out handling are enforced by the platform. What actually applies. This is not legal advice.

We recommend it, and most of our customers do. It costs almost nothing in outcomes and removes the worst failure mode — a person who feels deceived when they work it out mid-call.

Transcripts, summaries and ratings are stored against your workspace. Recordings have a retention window you configure. The detail is here.

Yes — the platform is available as a one-time licence for teams that need it inside their own environment. Talk to us about it.

Roughly 200 a month. Below that, the setup costs more than the calling saves and a person with a phone is genuinely the right answer. A four-question test.

Complaints and escalations, high-value negotiation, and anything legally sensitive — medical, legal or personal financial advice. Automate the reminder about the appointment, never the consultation.

No. Screening is one use case; order confirmation, lead qualification, appointment reminders, payment reminders and feedback are all bigger by volume. By industry.