Automate the clerical half. Keep the other one.

The test is simple — if the same input always deserves the same action, a machine should be doing it.

• Building a list from a description you can state, rather than from a spreadsheet somebody maintains by hand.
• Sorting that list, so the fifty most plausible businesses are at the top of five hundred.
• Sending on a schedule, in business hours, without anybody being awake.
• Following up when nobody replied, and stopping when somebody did.
• Retrying a step that failed for a reason that might not recur.
• Reading every incoming reply well enough to decide which ones a human should read first.

The same test in reverse. If the right action depends on something the system cannot see, automating it produces confident nonsense at volume.

• The claim you make about your product. A generated draft is a starting point; the claim is yours and you are accountable for it.
• Deciding a lead is worth pursuing. The score is a sort order built from a business listing, not knowledge of their budget.
• The conversation after a reply. This is the part the whole pipeline exists to produce.
• Consent, content and timing. Meta’s policies, TRAI’s hours and India’s data protection law all bind you, not the tool.
• Launching anything, if an AI proposed it. That is a human decision by design — see the approval gate.

Concretely, once a campaign is launched: it sends across WhatsApp, email, SMS and AI voice calls on a sequence you set, inside a send window that defaults to Monday to Friday, 9:30am–6:30pm IST. Scheduled steps live in a queue, so a step booked for tomorrow survives a restart and fires on time. Replies land in a shared inbox, classified, with a status, a label, an assignee and internal notes.

And a campaign whose channel is not properly configured refuses to launch rather than launching and failing one message at a time — which is the difference between finding out now and finding out from your quality rating.

The cluster pages have the detail: building the list, the WhatsApp rules, and what the AI is not allowed to do.

Will this replace an SDR?

No, and a tool that says otherwise is selling you something. It removes the list-building, the scheduling, the retry logic and the sorting of replies — the parts of the job that are clerical. The conversation that follows a reply is still the job.

Can I run it from my own systems instead of the dashboard?

Yes. The Automation API takes leads in from any system and runs outreach on them, and there is a WhatsApp send endpoint, both authenticated by a workspace API key.

What happens if a step fails?

Scheduled work lives in a queue rather than in a process. A step that fails is retried; a step scheduled for tomorrow’s send window survives a restart and fires on time. A campaign will also refuse to launch if the channel it needs is not actually configured, rather than launching and failing per message.