HVAC AI receptionist: missed-call and after-hours capture
What an HVAC AI receptionist does with the calls your office cannot answer: KPIs to demand, a demo script with traps, TCPA and PECR notes, and vendor questions.
By Voice Agent Bible Research · 4 min read
Last verified 30 Sept 2026v1.0Published 30 Sept 2026
KPIs at a glance
| KPI | Typical baseline | Target | How to measure |
|---|---|---|---|
| Answer rate, all hours | One call-analytics vendor's platform data puts unanswered inbound calls to home-services businesses at about 27%; count your own for one week | 99% of offered calls answered within three rings, including nights, weekends and weather spikes | Carrier or platform logs: answered / offered, by hour of day and by day, first 30 days. |
| Captured jobs from previously missed calls | Fewer than 3% of callers who reach voicemail leave a message (same source); assume near zero | A booked or dispatch-ready job on at least half of the calls that would previously have rung out | Jobs created outside staffed hours or during overflow / calls answered in those windows; value at average ticket. |
| Booking containment | Not applicable before deployment | 60-75% of booking intents end with a job on the dispatch board, no human involved | Calls ending with a written job in the field-service system / calls with a booking intent. |
| Address and callback accuracy | Message-taking services regularly deliver wrong or incomplete addresses; rarely counted | Under 1% of agent-created jobs edited by dispatch for address or phone within 24 hours | Jobs corrected for address, unit or phone / agent-created jobs, weekly. |
| Emergency hand-off time | Voicemail: hours. Answering service: minutes to a page, no context | On-call technician alerted with transcript and address within 60 seconds of the emergency being recognised | Timestamp of emergency phrase in transcript to timestamp of the on-call alert. |
| Voice-to-voice latency on availability turns | Rule of thumb used across this site: above about 1.2 s per turn the agent feels like an IVR | Median under 0.8 s; 90th percentile under 1.5 s on turns that read the dispatch board | End of caller speech to first agent audio, from recordings or platform traces, tool-backed turns only. |
What it is
A missed-call capture agent answers the calls your office cannot. It picks up when the line rings out during the day because everyone is on a job, when the office is closed, and when a heatwave or a freeze triples the volume. It works out what the caller needs, looks up the customer, checks real technician availability, books the job with an arrival window, and writes it to the dispatch board with the lead source attached. When the call is an emergency, it gives a safety instruction and alerts the on-call technician. When the call is out of scope, it says so and takes a message.
The call shape is short: greeting and disclosure, intent, address and callback capture with read-back, availability and window, confirmation, write, close. Two to three minutes is normal. Most of that is the caller talking, often while standing next to the broken unit.
Owner-operators buy this as an AI receptionist or an AI answering service for contractors. Multi-branch groups buy it as overflow and after-hours automation for dispatch. Same job, different volume, deeper integration.
Who buys it
- Owner-operators and office managers at one- to ten-truck HVAC, plumbing and electrical companies. The trigger is usually a week of counted missed calls, or one lost install job that a competitor booked at 7 p.m.
- Multi-branch home-services groups replacing a message-taking answering service that cannot see the schedule, or thinning a central call centre's night shift.
- Franchise operators who need the same script and the same booking rules across many locations and one field-service platform.
Budget owner: the owner or general manager. In groups, the call-centre or operations lead, with the field-service system administrator signing off on the integration.
KPIs
Count missed calls for one week before you deploy, by hour. That baseline is the whole business case. Then track the strip above: answer rate across all hours, jobs captured from calls that would have rung out, booking containment, address and callback accuracy, emergency hand-off time, and voice-to-voice latency on the turns that read the dispatch board.
Two measurement traps. "Captured" means a job on the board or a dispatch-ready record with a confirmed address and callback number, not a call that ended politely. And most missed calls in this trade happen during the working day, so a deployment that only covers nights is measuring the smaller half of the problem.
Demo script
Run the full script from the demo guide. The condensed version, with the traps that separate a product from a demo:
- Overflow entry and greeting. Call the main line and let it ring past the count you set. Pass: the agent answers, discloses that it is an AI, and gives the recording notice without being asked.
- Intent with interruption. Start describing an air conditioner that stopped, then cut in mid-offer with "actually it is only the upstairs unit". Pass: it stops talking and takes the new detail. Fail: it finishes its sentence and offers the old slot.
- Ambiguous address. Give "42 Oak" in a town that has an Oak Street and an Oak Avenue. Pass: it asks which, then reads the full address back with postcode or ZIP and asks about a unit number.
- Phone read-back. Say your callback number quickly. Pass: it reads the number back digit by digit and waits for a yes. Fail: it repeats the number as a single chunk or skips confirmation.
- Time window. Ask for "tomorrow afternoon". Pass: it names the date and offers the arrival windows the dispatch board actually has, such as twelve to four, rather than inventing a time.
- Eight seconds of silence after it offers a window. Pass: a short prompt, then a graceful hold or a message option. Fail: it hangs up, or repeats the whole offer.
- Emergency injection. Say "hang on, I think I can smell gas near the furnace". Pass: it drops the booking, tells you to leave the house and call the gas emergency line, and alerts the on-call technician in the same turn. Fail: it keeps offering windows.
- Out-of-scope request. Ask for a quote on a new roof. Pass: it says this is not a service you offer and asks whether to leave a message. Fail: it books a "roof" job.
- Confirmation and write. Accept a window and watch the job appear on the sandbox dispatch board with the address, window, job type, notes and lead source.
- System down. Have the sandbox connection disabled and call again. Pass: it says it is taking details, captures a full record and promises a callback within the time you configured. Fail: it says "you are booked".
Score each trap pass or fail. A vendor who wants to run the demo from their own audio has not passed the demo.
Compliance notes
This use case is inbound, and a call the customer initiates is outside the TCPA consent rule in the United States. Recording is the live issue: federal law is one-party consent but roughly a dozen states require all-party consent, so the agent should announce recording at the start of every call. If the agent later sends an SMS confirmation or places a callback, that is an outbound artificial-voice or text contact and needs prior express consent and the 8 a.m. to 9 p.m. window in the customer's local time. In the United Kingdom, a recording is processing of personal data under UK GDPR; the agent should say so at the start and the vendor should document the lawful basis. Any follow-up marketing call by the agent falls under PECR Regulation 19 and needs specific prior consent. In Australia, surveillance-devices laws differ by state and several require all-party consent, so announce recording; any outbound follow-up sits inside the telemarketing hours of Monday to Friday 9 a.m. to 8 p.m. and Saturday 9 a.m. to 5 p.m. and must respect the Do Not Call Register. The compliance rows for your regions are listed on this page. They are informational, not legal advice.
Build or buy
Buy a packaged product if you are a single-location contractor on a mainstream field-service platform; the integration and the overflow forwarding are the hard parts and vendors have done them. Consider a platform or a build if you run many locations with custom booking rules, zone logic or a membership tier that changes what the agent may promise. In both cases the acceptance test is the same: a job on your own sandbox dispatch board during the demo, and a gas-smell call that gets the safety instruction in one turn.
Questions to ask vendors
- 01
Show me a job landing on a sandbox dispatch board in our field-service system, with the right job type, address, arrival window and lead source.
A good answer: A job you can see on the board during the call. A lead that appears only in the vendor's dashboard, or an email to the office, is a message-taker.
- 02
How does the agent pick up overflow calls during the day as well as after hours, and who controls the ring count before it answers?
A good answer: Carrier-level forwarding on no-answer or busy with a ring count you set, plus a schedule you edit yourself, shown in a configuration screen.
- 03
How does the agent confirm the service address and the callback number it heard?
A good answer: Street name with type and postcode or ZIP read back, unit or apartment asked for explicitly, phone number read back digit by digit before the job is written.
- 04
What does the agent do when a caller says they smell gas, see sparks from the panel, or have water coming through a ceiling?
A good answer: Drops the booking flow, gives the scripted safety instruction (for gas: leave the house and call the gas emergency line), and alerts the on-call technician in the same turn. Shown in a transcript.
- 05
What happens when the caller asks for something we do not do, like a roof or a septic tank?
A good answer: It says plainly that this is not a service you offer, takes a message if you want one, and does not invent a booking.
- 06
What is the median and 90th-percentile voice-to-voice latency on a turn that reads availability, and how was it measured?
A good answer: Numbers for tool-backed turns, not greetings, with a method you can reproduce from your own phone.
- 07
What does the agent do when our field-service system is slow or down at 11 p.m.?
A good answer: It tells the caller it is taking details, captures a full structured record, and queues the write with an alert to the office. It never says 'booked' when nothing was written.
- 08
What is the all-in cost per connected minute for about 900 calls a month, including telephony, speech and the language model, and what does the month of the first heatwave cost?
A good answer: A line-item breakdown, a monthly total, and the number at three times normal volume.
Matrix rows that apply
Rows from the global compliance matrix that apply to this page. Informational only, not legal advice; dates change, confirm with counsel and the regulator.
| Jurisdiction | Consent for automated calls | AI disclosure | Calling hours | Recording | Verified |
|---|---|---|---|---|---|
| United States (federal)confidence high | Required The FCC's February 2024 declaratory ruling confirms that AI-generated or cloned voices are "artificial or prerecorded" voices under the TCPA. Outbound calls using them need prior express consent; marketing calls to mobile numbers need prior express written consent. Inbound calls initiated by the consumer are outside this consent rule. | Conditional No federal statute yet requires an agent to announce that it is AI. TCPA rules already require prerecorded or artificial-voice calls to identify the caller at the start and give a callback number. An FCC proposal (2024) would add an explicit AI disclosure; several states have their own bot-disclosure laws. Disclose by default. | Required Telephone solicitations only between 8 a.m. and 9 p.m. in the called party's local time (47 CFR 64.1200(c)(1)). | Conditional Federal law is one-party consent; roughly a dozen states (including California, Florida, Washington and Pennsylvania) require all-party consent. Announce recording at the start of every call unless counsel confirms otherwise. | 2026-09-30 |
| United Kingdomconfidence medium | Required The ICO treats conversational AI voice calls as automated calls under PECR Regulation 19, so direct marketing by automated call needs the recipient's specific prior consent. Live human marketing calls follow the softer Regulation 21 rules (screen against the TPS). | Recommended No UK statute mandates announcing an AI caller, but PECR requires automated marketing calls to identify the sender and provide a contact address, and UK GDPR transparency duties apply. | Recommended No statutory hours in PECR; Ofcom and industry codes expect reasonable hours and honouring "do not call again" requests. | Required Recording is processing of personal data under UK GDPR; tell callers at the start and document the lawful basis. Financial firms have additional FCA recording duties. | 2026-09-30 |
| Australiaconfidence medium | Required Telemarketing calls must not be made to numbers on the Do Not Call Register without consent (Do Not Call Register Act 2006); research calls have narrower exemptions. | Conditional The Telemarketing and Research Calls Industry Standard requires callers to identify themselves, the organisation and the purpose at the start. No general AI-caller law; broadcasting codes have begun requiring synthetic-voice disclosure in specific contexts. | Required Telemarketing calls only Monday to Friday 9 a.m. to 8 p.m. and Saturday 9 a.m. to 5 p.m. local time; none on Sundays or national public holidays (Industry Standard 2017). | Conditional State and territory surveillance-devices laws differ; several require all-party consent. Announce recording at the start. | 2026-09-30 |
| New Zealandconfidence low | Recommended No statutory do-not-call register for voice calls; the Marketing Association's Do Not Call list is voluntary. The Privacy Act 2020 governs collection and use of personal information. | Not required No AI-caller disclosure statute; Privacy Act transparency principles apply. | Recommended Industry code expectations only. | Recommended One-party consent for a participant; notify callers to satisfy Privacy Act collection principles. | 2026-09-30 |
Frequently asked
Is an HVAC AI receptionist the same as an after-hours answering service?
No. An answering service takes a message and pages someone. An AI receptionist reads the dispatch board, books the job with an arrival window, and pages a person only for emergencies and exceptions. Vendors use both names for both products, so ask for a live write into your field-service system.
Can the agent answer overflow during the day, not only after hours?
Yes, if your carrier or phone system forwards on no-answer or busy after a ring count you choose. Most missed calls in this trade happen during the working day while technicians are on jobs, so daytime overflow is usually worth more than the night shift.
What does the agent do with a gas smell or a flood?
A well-built agent recognises the phrase in one turn, gives a scripted safety instruction (leave the house and call the gas emergency line for gas; shut the main water valve if it is safe for a flood), and alerts the on-call technician with the address. Test it with your own recorded call before you sign.
How much does an AI receptionist for a contractor cost?
Small-business products are usually monthly subscriptions with a minute allowance; usage-priced platforms bill per connected minute plus telephony. Advertised per-minute prices commonly exclude speech, language-model and telephony costs. The TCO calculator has a home-services missed-call preset.
Related
- Home services #2
- Home services #3
- Home services #4
- Home services #5
- Demo script
- Best practice
- Best practice
- Best practice
- Anti-pattern
- Anti-pattern
- Anti-pattern
- Tool
- Market
- Market