I built an AI chatbot for my dad's plumbing business. Here's what it cost.
A real case study with real numbers. What it took to go from 'my dad hates missing calls' to a working booking assistant.
My dad runs a two-person plumbing outfit. He misses roughly a third of the calls that come in because he's under a sink. Every missed call is roughly $180 of average job value that walks to a competitor.
I decided to fix it with an AI booking assistant. Here's exactly what it cost and how long it took.
The stack I ended up with
- Twilio for the phone number and voice routing
- OpenAI Realtime API for the voice model
- A tiny web app (built on Lovable) for the booking calendar my dad checks in the morning
- Google Calendar for the actual bookings
Total moving parts: 4. Fewer than I feared.
The bill for the first month
- Twilio phone number + minutes: $14.20
- OpenAI Realtime usage (about 180 minutes of calls): $47.10
- Lovable Pro: $25.00
- My dad's coffee while I explained it to him five times: priceless
Total: ~$86 for the month.
The bot booked 11 jobs he would have missed. At his average ticket, that's about $2,000 in captured revenue.
What I got wrong the first time
- Made the greeting too long. Callers hung up. Cut it to one sentence.
- Let it answer pricing questions. It made things up. Now it says "Frank will call you back with a quote."
- Didn't test it with his actual accent. The bot kept correcting his name. Fixed by giving it his name in the system prompt.
What actually mattered
The killer feature wasn't the AI. It was the SMS transcript the bot sends my dad after each call. He reads it between jobs, calls the customer back if it feels off, and stays in control. Nobody wants a bot to replace the human — they want it to buy the human 30 more minutes a day.
Would I do it for someone else?
Yes. And I'd charge $2,000 to build it, $150/month to maintain it. If you're reading this and you own a small service business, you can literally do this yourself in a weekend now. That's the wild part.
