Same truck, same prices. The only change was replying first.
Answer missed and after-hours calls (and book the job)
A complete after-hours call script and the instructions to paste into an AI receptionist (or give to an answering service).
Who it's for
Owners whose calls go to voicemail while they're on a job, at lunch, or after 5.
What you'll have at the end
A complete after-hours call script and the instructions to paste into an AI receptionist (or give to an answering service).
Time and cost
About 1 hour to set up the script. AI receptionists run from about $29/mo (Reception.ai Basic) to a few hundred a month; see the Tools page. Time saved: estimate only, varies with call volume.
Before you start
- A ChatGPT or Claude account (free works)
- your hours, services, rough price ranges, and your emergency rules
- the on-call phone number
Step-by-step
Write down what counts as an emergency for you (e.g. no heat with kids in the house).
The prompt
0 copiesFill these in
| Field | What it means | Example |
|---|---|---|
| [BUSINESS NAME] | your shop's name as customers know it | Lakeside Auto Care |
| [CITY] | where you're located; used for local details | Oconomowoc, WI |
| [TONE] | how you talk to customers | friendly, straight-talking, no jargon |
| [SERVICES] | what you do | brakes, diagnostics, oil changes, tires |
| [HOURS] | your open hours | Mon-Fri 7-5 |
| [EMERGENCY RULES] | what counts as urgent and who gets the call | tow-ins: take info, text owner at 414-555-0100 |
| [BOOKING] | how jobs get booked | next open slot in the shop calendar |
How much are missed calls costing you? Try the missed-call calculator
Want help getting this working? Rise Systems in Oconomowoc sets these up.
Follow-ups
What good output looks like
Short questions, one at a time; a clear split between urgent and routine; no prices promised; the caller always knows what happens next.
If it goes wrong
- It sounds robotic: ask for 'shorter sentences, like you're talking across the counter.'
- It promises prices: add 'never quote prices, say a tech will confirm' to the rules.
- It's too long for phone use: ask 'cut to the 5 questions that matter most.'
- It misses your emergencies: paste 3 real past calls and ask it to adjust.
Level up
Turn on an AI receptionist (Reception.ai, Smith.ai, or Jobber/Housecall Pro's built-in AI) and paste the rules in. Pros: answers 24/7 and books into your calendar. Cons: needs a week of tuning; texting needs A2P 10DLC registration; plans have minute or call caps.
Worth knowing
Test it by calling it yourself before customers do. Make sure real emergencies still reach a human.
Wisconsin angle
Cold-snap breakdowns, deer collisions, and Packers Sundays are when calls pile up and nobody's at the desk.
Source: Wisconsin Trades AI editorial
A counter-story from the customer's side: the bot didn't know the price of the company's own maintenance plan.
Fixed price list, a zip-code check, preset time slots. Anything else becomes a message.
Worth knowing before you pay big for setup: your software may already include it.
Calls get answered when everyone's in the bays, and he gets summaries. Some people still want a live person.
Voicemails weren't getting checked. Now after-hours calls become lead and appointment requests.
Missed calls dropped a lot. Customers could tell it was AI within a sentence or two.
It never forgets to ask questions the office might. Some customers have complained to him about it on site.
'Still a personable and human company,' and still catching the leads it used to miss.
The winner booked paid jobs overnight on its own but couldn't remember past callers.
It caught a duplex with two driveways and a pool deck at night: booked, deposit paid, done that Saturday.
The customer's side, and positive this time: a bit of uncanny valley, but a good experience.
Humans still answer first; AI takes overflow and captures just 4 things.
An honest counter-story: fine for basics, but it fell apart on frustrated callers.
Customers walk in sure of their diagnosis because ChatGPT told them. Sometimes it's dangerous.
A real win with an asterisk: it took a lot of setup to get there.
One said Bland took about two weeks of tuning. Another said multi-topic callers broke theirs.
Quotes that took him an hour per 10 now come out with datasheets in about 15 minutes.
The best answer in the thread: use AI to write tools that do the work, not to do the work itself.
One customer's ChatGPT said a new condenser fan motor was wrong and recommended a blower motor instead.