AI Receptionist for HVAC and Plumbing Companies: Setup Guide, Call Script, and Emergency Rules

If you run an HVAC or plumbing company, your phone is your cash register. The problem is that the people who should answer it are usually on a roof, in a crawlspace, or driving between jobs. After hours, the phone goes to voicemail, and the caller with a burst pipe tries the next company.

An AI receptionist (also called an AI voice agent or AI answering service) answers those calls, has a natural back-and-forth conversation, captures the job details, and either books the appointment or escalates real emergencies to your on-call tech.

This guide is for owners and the agencies that serve them. It covers what an AI receptionist should and shouldn't do for a trades business, a practical call flow, emergency triage rules, a sample prompt, a testing checklist, and what it actually costs to run in HighLevel Voice AI, using HighLevel's published rates as of October 2026. For a standalone, copy-ready overnight script with gas/CO safety lines and a four-level triage table, see the HVAC after-hours answering script.

What should an AI receptionist do for an HVAC or plumbing company?

It should answer every call, capture the job details, book routine work, and escalate emergencies. A well-configured AI receptionist for HVAC or plumbing handles the predictable, routine calls so your team can focus on the rest. Realistic jobs for it:

• Answer every call, including overlapping ones, after hours or as overflow during the day.
• Capture the job details: name, callback number, service address, ZIP, and what's wrong.
• Sort the call: emergency, repair, maintenance or tune-up, estimate for a replacement, or existing job question.
• Book appointments on your calendar for routine work.
• Escalate emergencies by transferring the call or alerting the on-call tech immediately.
• Answer common questions from an approved knowledge base: service area, hours, whether you charge a trip fee, brands you service, financing availability.
• Send a text confirmation with what happens next.

What should an AI receptionist never do?

It should never diagnose, quote prices you haven't approved, improvise safety advice, guess, or pretend to be human. This is where most bad setups go wrong. Write these limits into the prompt explicitly:

• No diagnosing. "It sounds like your capacitor" is a promise you'll have to walk back. The AI collects symptoms; the tech diagnoses.
• No firm prices unless you have flat-rate items you're happy to quote. Ranges or "the tech will quote on site" are safer.
• No safety advice beyond your script. For gas, carbon monoxide, or electrical hazards, the AI should give one short, pre-approved instruction and escalate, not improvise.
• No guessing. If the answer isn't in the knowledge base, it takes a message. HighLevel's own setup guide makes the same point: tell the agent not to guess when required information is unavailable (HighLevel: Create, Test, and Deploy a Voice AI Agent).
• No pretending to be human. Have it introduce itself as a virtual or AI assistant. Callers handle that better than discovering it mid-call.

What call flow works for HVAC and plumbing calls?

A short, predictable flow that checks for safety first, then collects the caller's details, then books or escalates. Here's a flow you can map almost directly onto a prompt:

1. Greeting and disclosure. "Thanks for calling [Company], this is the virtual assistant. Are you calling about an emergency, or would you like to schedule service?"
2. Safety check first. If the caller mentions gas smell, a CO alarm, smoke, sparks, or flooding near electrical, jump to the emergency script (below).
3. Identify the caller. Name and best callback number. Confirm the number by reading it back.
4. Service address and ZIP. Check it against your service area. If it's outside, say so politely and offer a referral if you have one.
5. The problem. For HVAC: heating or cooling, what it's doing (not running, running but not cooling, leaking, noise), system type if they know it. For plumbing: leak, clog, water heater, sewer smell, no water, and whether water is actively spreading.
6. Urgency. Is anyone vulnerable at home (infants, elderly, medical needs)? Is it extreme heat or freezing weather? This decides whether it's an emergency under your policy.
7. Book or escalate. Routine work gets the next available slot. Emergencies go to the on-call tech per your rules.
8. Recap and close. Read back the appointment or next step, mention any trip or after-hours fee if you charge one, and confirm a text is on the way.

How should an AI receptionist handle HVAC and plumbing emergencies?

It should follow a written triage table: safety hazards get a short safety line and an immediate escalation, and everything else gets booked by urgency. The AI can only follow rules you've written down. Agree on these with your team first, then paste them into the prompt word for word.

The safety lines above are common guidance, but follow your own company policy and local utility instructions. Keep them short. A caller smelling gas doesn't need a paragraph.

What information should the AI capture on every call?

Capture who is calling, where the job is, what's wrong, and how urgent it is. Make these required fields so every job lands in your CRM complete:

• Full name, callback number (confirmed), email if they offer it
• Service address and ZIP
• Homeowner, renter, or property manager (renters may need landlord approval)
• Service type: HVAC heating, HVAC cooling, plumbing, water heater, drain, other
• Description in the caller's own words
• Urgency category from your triage table
• System details if known: age, brand, gas or electric, tank or tankless
• Preferred appointment windows
• How they heard about you (useful for marketing, optional)

In HighLevel Voice AI, the Update Contact Field action saves information collected from the caller to the contact record, and Trigger a Workflow can start your dispatch or notification automation after the call (HighLevel docs).

What does a good HVAC and plumbing AI receptionist prompt look like?

It defines the role, the goal, safety-first rules, hard limits, and how to close. Here's a condensed version you can adapt. The full HVAC and plumbing prompts, with knowledge-base templates, are in our free Prompt Pack.

ROLE: You are the virtual receptionist for [Company], an HVAC and plumbing
company serving [cities/ZIPs]. You answer calls warmly and briefly.
Always say you are a virtual assistant in your greeting.

GOAL: Capture name, callback number, service address, ZIP, problem, and
urgency. Book routine jobs on the calendar. Escalate emergencies.

SAFETY FIRST: If the caller mentions a gas smell, CO alarm, smoke, sparks,
or water near electrical, stop intake. Say: "[approved safety line]".
Then run the Emergency Transfer action.

RULES:
- Never diagnose or promise a cause. Say a technician will assess on site.
- Only quote prices listed in the knowledge base. Otherwise: "The
  technician will give you an exact price before any work starts."
- If you don't know, say you'll have the team call back. Never guess.
- One question at a time. Confirm phone numbers and addresses by
  reading them back.
- Service area: [list]. Outside it, apologize and offer [referral/none].

CLOSE: Recap the appointment or next step, mention [after-hours fee policy],
and tell the caller a text confirmation is coming.

How do you build it in HighLevel Voice AI?

You create an agent in the sub-account, add a knowledge base and actions, then deploy it to a phone number with your working hours. HighLevel's Voice AI lives inside each sub-account under AI Agents → Voice AI. Per HighLevel's help docs (October 2026), you can start from a Marketplace agent, build a Custom Agent with a prompt, or use the Flow Based Builder for structured, multi-step conversations. The Flow Builder also has templates such as Front Desk Receptionist, Business FAQ Assistant, and Lead Qualification (HighLevel: Voice AI Flow Builder templates).

The setup, in order:

1. Core settings. Agent name, voice, model, language, time zone, greeting, and the prompt.
2. Knowledge base. Attach one with your service area, hours, brands, fees, financing, and FAQs. HighLevel lets you describe when the agent should search it (HighLevel: Knowledge Base for Voice AI).
3. Actions. The ones that matter most for trades: - Appointment Booking, connected to a HighLevel calendar. It can book on a single calendar or route across multiple calendars by intent, such as repair vs. estimate (HighLevel: Appointment Booking in Voice AI). - Call Transfer to your on-call number for emergencies. - Update Contact Field for address, system type, urgency. - Send SMS for confirmations. - Trigger a Workflow to notify dispatch or create an opportunity. - Custom Action (a webhook) if you need to push jobs into field-service software.
4. Deployment. Assign a phone number, then choose a routing mode: Answer Calls Directly or Use as Backup, where the agent only answers when the primary call path doesn't. Set Working Hours to all hours or a custom schedule, for example evenings and weekends only.

A common pattern for small shops: keep the office line ringing to staff during business hours, use the agent as backup when nobody picks up, and let it answer directly after hours. Add missed call text back as a safety net for calls that still slip through. Plumbing companies can use our missed call text back scripts for plumbers, which include leak-safety reply lines and an URGENT alert workflow.

How should you test an AI receptionist before launch?

Test it like a difficult customer, with real phone calls as well as browser tests. Don't go live after one friendly test call. HighLevel's docs recommend testing common requests, questions the agent should not answer, unclear requests, each action, transfers, silence, and interruptions. Note that call transfers don't run in browser "Web Call" tests, so test them with a real phone call.

Run at least these scenarios before launch:

• A calm tune-up booking
• An upset caller with no AC in a heat wave
• A gas-smell call (confirm it stops intake and escalates)
• A caller outside your service area
• Someone asking "how much to replace my water heater?"
• A caller who gives the address in pieces, or changes it mid-call
• Background noise, long silence, and talking over the agent
• A spam or robocall

Then read the transcripts and call summaries. HighLevel saves call logs with transcripts and summaries, and you can choose whether summaries are also written to the contact's notes.

How much does an AI receptionist cost to run in HighLevel?

On pay-per-use, the AI usage is about $0.06 per minute (roughly $18 for 300 minutes) before tokens, plus the platform plan and phone charges. HighLevel prices Voice AI separately from its platform subscription. As of October 2026, from the HighLevel pricing page and AI Product Pricing docs:

• Platform plan: Starter $97/month, Unlimited $297/month, Pro $497/month (annual: $970, $2,970, $4,970).
• Voice AI options per sub-account: Pay-per-use (no monthly AI fee), AI Employee Growth at $50/month (includes 100 Voice AI minutes), or AI Employee Unlimited at $97/month (unlimited Voice AI, subject to fair use).
• Pay-per-use Voice AI rate: Voice Engine $0.045/min + text-to-speech (e.g., $0.015/min for OpenAI or Cartesia voices, $0.170/min for ElevenLabs V3) + LLM token cost.
• Phone system charges apply on every plan, including Unlimited. A US local number is $1.15/month, and US inbound call minutes are billed at phone-system rates (for example, $0.01165/min for inbound to web, mobile, or desk phone in HighLevel's guide).

Here's an illustration of AI charges only, for an agent using a $0.015/min voice, before LLM tokens and phone charges:

The math is simple: minutes × ($0.045 + $0.015). For most small shops answering after-hours and overflow calls, pay-per-use is likely the cheapest of the three. The flat plans make sense at high volume or when an agency wants predictable costs. Rates change, so check HighLevel's AI pricing page and your in-app billing before you quote anyone.

What are the alternatives to HighLevel?

The main alternatives are standalone AI receptionist apps, human or hybrid answering services, and your field-service software's own tools. HighLevel isn't the only route. Broadly, your choices are:

• An all-in-one platform like HighLevel. Phone system, AI agent, CRM, calendars, text-back, and workflows in one place. More setup, but everything connects. Strong fit for agencies serving several trades clients.
• A standalone AI receptionist app. Usually faster to start and built around one job. You may need integrations to get leads into your CRM or field-service software.
• A human answering service, or a hybrid AI-plus-human service. Higher cost per call, but a person handles the tricky calls.
• Your field-service software's built-in tools. Some platforms used by trades include call or text features. Check what you already pay for.

We're an affiliate of HighLevel, so read our comparison with that in mind. We are not speaking for HighLevel, and we suggest trying a couple of options on real after-hours calls before you commit.

FAQ

Will customers hang up on an AI? Some will. Clear disclosure, a fast greeting, short questions, and an easy path to a human help. Review transcripts weekly and adjust.

Can it transfer to my cell phone? Yes. In HighLevel, the Call Transfer action sends the caller to a human phone number under conditions you define.

Does it work with ServiceTitan, Jobber, or Housecall Pro? HighLevel Voice AI can send data to outside systems through a Custom Action (webhook) or workflows. Whether a direct integration exists for your software is something to confirm with HighLevel and your software vendor before you buy.

How long does setup take? A basic agent can be built in an afternoon. Getting the triage rules, knowledge base, and transfers right usually takes a few rounds of testing.

Sources (checked October 3, 2026)