The rules it runs under.
It identifies itself in the first sentence. It asks what's wrong before it asks when. It never quotes a price unless you've given it an exact rule, and then it says only that. It hands off the moment a caller asks for a person. It never promises the work — no "we can definitely fix that," no timelines it wasn't given. Every call becomes a record with a transcript, so nothing it says is ever a mystery to you.
The setup.
A couple of hours with us on how you sell, what you charge, and what the agent may and may not say. We write its rules from that conversation. Then we call it — repeatedly, in every way a customer might, at every hour — before a customer ever reaches it. Early on we read the transcripts and tune what it says from real calls, not assumptions.
When it fails.
It will sometimes misunderstand; anything that listens to speech does. What matters is what happens next. You see the transcript either way. We'd rather you read a mistake in a log than lose a caller to voicemail.
Where the recordings live.
Your calls are your customers' voices. We've handled other people's data under security obligations for ten years at Watchman. Your call recordings get treated like a client's server.
What it's up against.
Voicemail — where roughly nobody leaves a message anymore, and the cost never shows up on an invoice. An answering service — a person reading a script who can't see your calendar, converting a live caller into a message that's already cooling. The thirty-dollar AI receptionist — a chatbot with a voice that can't reach your calendar, creates no record, and has nobody watching it at 2 AM. What we run is none of these: your rules, your calendar, your CRM behind it, and three people reading the transcripts.
What it doesn't do.
No outbound cold calling. No pretending to be human. No quoting without your rules. No legal or medical advice. It doesn't replace a good front desk during the day — it covers the hours and the overflow they can't.