Smith.ai

A prospect submits an intake form in Typeform or Gravity Forms. The agent enriches the submission from the CRM, works out whether this is a new matter or an existing client, and drafts a short brief for the receptionist covering who they are, what they asked about and what to confirm. Request Outbound Call carries that brief as the instructions, plus fallback instructions for a voicemail, and an availability setting of 24/7 or Custom with its retry cadence and attempt count. The person who filled in the form at 9pm gets a call from a human, and the firm did not need someone sitting by the queue to make that happen.

Rescheduling is a task automation handles badly, because the second message in the exchange is almost always a counter-offer. The agent detects the trigger, whether that is a cancellation, a conflict in Google Calendar or a no-show, and hands the whole thing to a receptionist with instructions to book or move the consultation. Smith.ai's receptionists can also cover a bilingual conversation where the caller's preferred language is not English, which is stated in the instructions rather than configured as a flag. The outcome comes back through the post-call webhook and the agent writes it to the record.

New contacts arriving in the CRM can be fed into an existing Smith.ai Outreach Campaign with Add Recipient to Outreach Campaign, identified by the campaign UUID from its dashboard URL. Up to three user data fields ride along, invisible to the calling agent but present in the post-call payload, which is how a call gets reconciled back to the record that produced it. The agent does not add contacts on its own judgment, though. It proposes the batch with the source of each contact and the consent state on the record, and a person confirms who is in scope. Adding someone to a calling sequence is a decision about whether they should be called at all, and that is not a classification problem.