How a two-person plumbing business stopped losing $4k a month to voicemail

Two plumbers, one van each, phones on silent from the first job of the morning. Their week had a familiar shape: dig through voicemail at 6pm, call back the five numbers that left messages, discover two had already booked someone else, and never know how many hung up without leaving anything.

When they ran the numbers, the picture was blunt: roughly ten missed calls a week, an average job around $380, and — being generous — half of those callers gone for good. Call it $4,000 a month walking to competitors, for a business whose actual work had a five-star reputation.

The setup

The fix took one afternoon. They kept their existing number and forwarded it — no new number on the van, no reprinted cards. An AI receptionist now answers every call in under a second, knows their services and callout fees, and books jobs directly into a shared calendar with travel time between suburbs baked in. Emergencies — burst pipes, no hot water — ring straight through to whichever plumber is on call.

The part they did not expect to matter most: the after-call texts. Every booked customer gets an instant confirmation, and every quote gets followed up automatically two days later. The "phone tag" hours between 6pm and 8pm simply stopped existing.

Ninety days later

Booked jobs from inbound calls rose by a third — not because more people called, but because the ones who called got answered. The owners' own measure is simpler: voicemail is empty, evenings are theirs, and the only calls they personally take are the ones worth interrupting a job for. The $4k leak did not shrink. It closed.

The objections they almost stopped for

"Our customers want to talk to us, not a robot." That was the first worry, and it deserved a real answer. What they found in practice: customers want their problem handled. The caller with water coming through the ceiling does not care who answers — they care that someone answers now, knows what a tempering valve is, and can say "he can be there at 2pm". The AI saying honestly that it is the office assistant, then solving the problem, beat the voicemail every single time.

"What if it books something wrong?" It did, early on — twice in the first fortnight, both times a job type it had not been taught. Each mistake became a taught rule the same day ("no gas work bookings — those need a phone call with Dave first"), and the error rate went to roughly zero. The lesson generalises: an AI receptionist is not configured once, it is coached briefly and then it stays coached.