Home services businesses (HVAC, plumbing, roofing, electrical, lawn care) have specific setup requirements: emergency triage rules, dispatch logic, service-area filtering, and integrations with field-service platforms like ServiceTitan, FieldEdge, and Housecall Pro. A generic AI receptionist setup misses 30–40% of home-services callers — they ask "do you serve my ZIP", "how soon can you come out", "is this covered under warranty", and "can you take card on-site" — and a setup tuned to home services answers all four. This guide covers the home-services-specific configuration: call types, triage, dispatch, integrations, pricing rules, follow-up SMS, and what to monitor in week one.
1. Home-services call types you'll actually see
Tag every call into one of these buckets — they each have different rules:
| Call type | Volume | AI action |
|---|---|---|
| Service request (new) | 40–55% | Book or dispatch |
| Service request (repeat) | 15–25% | Look up customer, book |
| Emergency | 5–15% | Triage + escalate |
| Estimate/quote | 10–20% | Capture for callback |
| Status check | 5–10% | Look up job, give ETA |
| Billing question | 3–5% | Transfer or message |
| Sales/spam | 2–5% | Polite hang-up |
If your AI doesn't distinguish these, every call gets the same treatment — which means estimates get force-booked and emergencies get queued behind oil changes. The way we configure this for HVAC companies is the same template plumbers and electricians use.
2. Emergency triage rules
Every trade has its own emergency keywords. Set them explicitly in the script:
HVAC EMERGENCIES:
"no heat" + outdoor temp under 35°F
"no AC" + outdoor temp over 90°F + elderly/infant in home
"gas smell" → leave the house, call 911 first, then us
"carbon monoxide alarm"
PLUMBING EMERGENCIES:
"active leak", "flooding", "water everywhere"
"no water in the whole house"
"sewage backup"
"gas smell" → 911 first
"broken pipe"
ELECTRICAL EMERGENCIES:
"sparks", "burning smell"
"power out and someone on oxygen"
"wires exposed"
"shock hazard"
ROOFING EMERGENCIES:
"active leak through ceiling"
"tree on the roof"
"missing shingles after a storm" (ongoing weather)
LAWN/IRRIGATION EMERGENCIES:
"geyser" / "broken main"
"active leak from sprinkler"
Triage logic: emergency = SMS owner immediately, give caller an ETA, log to dispatch.
3. Dispatch routing
If you have more than one tech, the AI needs to know who gets what:
DISPATCH RULES:
- By skill: only assign drain jobs to techs flagged "drain-certified"
- By zone: ZIP-code-to-tech mapping (3 zones, 3 techs)
- By urgency: emergencies route to nearest tech regardless of zone
- By time: after 5pm only on-call techs get paged
ON-CALL ROTATION:
Mon-Wed: Mike
Thu-Sat: Sarah
Sun: rotates weekly
AI references: shared Google Calendar "On Call"
Most modern AI receptionists either read this from your dispatch software (ServiceTitan, FieldEdge, Housecall Pro) or from a shared calendar. If yours can't, it's a manual rule list — and you'll need to update it every time the rotation changes.
4. Service-area filtering
This single rule kills the most wasted calls: dispatching a tech 60 miles away for a $200 job.
SERVICE AREA:
Primary (free trip charge): ZIPs 14620, 14618, 14610, 14534
Extended (trip charge $89): ZIPs 14580, 14624, 14625
Out of area: politely decline, offer referral if available
ASK ZIP EARLY:
Question 3 of 4 must be ZIP. If out-of-area, exit politely:
"We don't serve [city] right now — sorry I couldn't help.
Try [referral company name + number]."
Filtering at qualification saves an average of 8–15 minutes per misrouted call [VERIFY] and avoids the "but the receptionist said you'd come out" complaint.
5. Integrations setup
The right integration depends on your field-service platform.
ServiceTitan
ServiceTitan exposes a public API for customers, jobs, and bookings. Common integration pattern:
- AI looks up caller by phone → returns customer record + service history
- AI creates a new job/lead in ServiceTitan with notes from the call
- Booking writes directly into the dispatch board
- Recording + transcript attached to the job record
Your vendor either supports ServiceTitan natively or via a middleware (Make, Zapier). Native is real-time; middleware adds a 1–5 minute delay.
Housecall Pro
Housecall Pro has a public API as of 2024. Pattern:
- Customer lookup by phone
- Job creation with category, address, notes
- Calendar push to assigned tech
- SMS confirmation triggers from Housecall, not the AI (avoids double-texting)
Jobber
Jobber has a well-documented API and webhook system:
- Client/property lookup
- Request → quote → job pipeline supported
- Most AI receptionists land calls in "Requests," then a human converts to quote/job
FieldEdge
FieldEdge integration is typically lighter — lookup + note creation. Booking usually still requires a human dispatcher confirming the slot.
What to test on day one
- Inbound call → customer record found (existing customer)
- Inbound call → new lead created (new customer)
- Booking lands on the right tech's calendar
- Notes from the call appear on the job record
- No duplicate customers created (most common bug)
6. Pricing and quoting rules
Home services pricing is messy. Set strict guardrails:
THE AI MAY QUOTE:
- Diagnostic / service-call fees ($79, $89, $99 — your number)
- "Starting at" prices for top SKUs (drain cleaning starts at $189)
- Trip charges by zone
THE AI MAY NOT QUOTE:
- Repair pricing
- Install pricing
- Replacement equipment
- Anything the tech needs to see in person
- Warranty coverage decisions
WHEN ASKED FOR A NUMBER ON FORBIDDEN ITEMS:
"Pricing depends on a few things — the tech will give you a
firm number on-site before any work. Diagnostic visit is $89,
waived if you proceed with the repair."
The "waived if you proceed" language is the highest-converting line in home services [VERIFY]. Worth including verbatim.
7. Post-call follow-up SMS
Most home-services calls end with a booking but no recap. SMS within 60 seconds:
TEMPLATE (booking confirmed):
"Hi {first_name}, this is {company}. We've got you on the
schedule {date} between {start_time}–{end_time}. Tech: {tech_name}.
Address: {address}. Reply C to confirm or R to reschedule."
TEMPLATE (estimate callback):
"Hi {first_name}, thanks for calling {company}. We have your
request and a team member will call within {window} to talk
about {service_type}. Reply STOP to opt out."
TEMPLATE (emergency dispatched):
"Hi {first_name}, {tech_name} is on the way for your
{issue_type}. ETA: {eta_minutes} min. Reply if anything changes."
Two-way SMS lets customers reply C/R/STOP. Don't send the AI's transcript — send the structured confirmation.
8. Monitor the first week
For the first 7 days, block 30 minutes daily to review:
- Booking rate (target: 40–55% of qualified callers)
- Escalation rate (target: 10–20%; higher means rules are too strict)
- Misrouted calls (out-of-area bookings — should be near zero)
- Duplicate customer records (most common integration bug)
- Calls > 5 minutes (AI got stuck — listen and fix)
- Calls under 30 seconds with no booking (caller hung up — usually greeting issue or carrier problem)
Most fixes are 60-second dashboard updates: adding a service, expanding the service area, fixing a tech name, adjusting an escalation trigger. By day 7 you should have a script that needs only weekly tweaks.
Frequently asked questions
Will an AI receptionist work with ServiceTitan?
Yes — ServiceTitan is one of the most commonly supported integrations among AI receptionist vendors. The standard integration covers customer lookup by phone, new lead/job creation with call notes, booking onto the dispatch board, and attachment of the call recording and transcript to the job record. A few caveats: ServiceTitan's API has rate limits, so high-volume shops should confirm their vendor handles backoff and retry. Some vendors integrate via the official API, others via middleware (Make/Zapier) which adds a 1–5 minute delay. Native is preferable for emergency dispatch. Also confirm the integration creates jobs in the correct business unit and campaign — getting that wrong scrambles your reporting for months.
How does the AI handle "do you take same-day appointments"?
Configure the AI to read your real availability from the dispatch board, not a static answer. If today's slots are open, it offers them. If not, it offers the next available — without saying "we're booked." The exact phrase that works: "I've got tomorrow at 9am or Thursday at 2pm — which works better?" That gives the caller a choice and avoids the open-ended "when can you come out" loop. For emergencies, override availability rules — the AI should always SMS the on-call tech regardless of the calendar. If you have a "rush fee" for same-day non-emergency work, tell the AI to mention it before booking ("Same-day is $50 extra — still want it?"). About 60% of callers say yes [VERIFY], which is pure margin.
Can the AI quote prices for repairs?
No — and you should explicitly forbid it. Repair pricing in home services depends on diagnosis, parts availability, equipment age, accessibility, and warranty status. An AI that quotes a confident number is going to be wrong, and customers remember the quoted number, not the on-site number. The right pattern: AI quotes the diagnostic fee ($79–$99), explains it's waived if work is approved, and gives the tech the conversation context. Some shops set a "starting at" price for common SKUs ("drain cleaning starts at $189") which is fine because the tech can adjust on-site. The AI should refuse with: "Pricing depends on what we find — the tech will give you a firm number before any work."
What happens with after-hours emergency calls?
The AI handles them 24/7 with a defined emergency script. After-hours flow: AI answers, qualifies the caller, identifies emergency keywords, captures name + callback + address + problem description, then SMS-pages the on-call tech with all of it pre-formatted. The caller hears: "[Tech name] is on call tonight — they'll call you within 15 minutes." If the on-call tech doesn't respond in 10 minutes, the AI auto-escalates to the backup. Non-emergency after-hours calls book into the next business day's calendar — without ever saying "we're closed." This setup typically recovers 30–50% of after-hours revenue that would otherwise go to voicemail [VERIFY], which for most home-services shops is the single biggest ROI driver.
How do I prevent the AI from booking jobs we don't do?
Maintain an explicit service catalog in the AI's knowledge base. List every service you DO offer with category and pricing rules. Add an explicit "DO NOT BOOK" list for adjacent work you don't take — common ones for HVAC: ductwork cleaning, dryer vent cleaning, indoor air quality testing. The AI should respond with a polite decline and (if you have one) a referral. "We don't do duct cleaning ourselves — Acme Duct usually handles that, their number is 555-1234." Referral exchanges build goodwill with neighboring businesses and stop the AI from making promises you can't keep. Review the "do not book" list quarterly — services creep in and out of your offering as you hire and lose techs.
Ready to see what an AI receptionist looks like for your business? → Book a 20-minute demo