$38,674 generated by the voice agent in one 30-day window

A plumbing company's revenue arrives by phone. When the office is slammed or the day is over, callers hit voicemail, and plumbing customers do not leave voicemails. A burst pipe at 9 PM gets fixed by whoever answers, so every missed call was quietly funding a competitor. Ironclad knew jobs were leaking. What was missing was a way to answer every call without hiring a night receptionist.

We put an AI voice agent on Ironclad's existing phone number. Nothing above it changed: same number, same branding, same crew. When a call comes in, the agent answers, asks what the caller needs, and classifies the job as routine, urgent, emergency, or general inquiry. It captures the address and contact details and books the visit into the calendar. A true emergency escalates straight to the owner's cell.

Every call lands in a searchable log with a transcript, an urgency tag, and a next step, so the office opens each morning to a queue of decisions instead of a mystery. The build runs on Twilio for telephony, Claude for the conversation and reasoning, ElevenLabs for the voice, and Supabase for the call log.

In its most recent 30-day window the agent answered 39 calls, converted 56.4 percent of them into booked work, and generated $38,674 in revenue, with a meaningful share of those calls arriving at hours no office answers.

The same pipeline keeps running today: every call answered, qualified, and either booked or escalated, around the clock.