Your phone is the front door. The whole crew is on a roof, under a sink, or in a crawlspace. So the front door is unattended for most of the working day. That is not a staffing problem you can hire your way out of cheaply. It is a structural leak, and it is measured in dollars per ring.
Home-service companies miss up to 62% of inbound calls, and across plumbing, electrical, and HVAC the miss rate runs roughly 27% to 62% because the technician answering the phone is the same person holding the wrench. The call does not wait. The customer dials the next name on the list.
The AI receptionist for service businesses that books the job while your crew is on site
This is not a generic small-business pitch. This is for operators who dispatch trucks. HVAC, plumbing, electrical, cleaning, repair, recurring-service. The kind of shop where the phone rings during a job and nobody can pick it up without stopping work. If that is you, read the math first. Then decide.
Lead with the leak: what the missed call actually costs
Start with the only number that matters: the cost of doing nothing. A missed call in the trades is not a missed conversation. It is a missed job, and the next vendor takes it.
Each missed service call is worth roughly $275 to $1,200 depending on the trade. A drain clear sits at the low end. A full HVAC system replacement sits at the high end. A contractor missing 5 to 10 calls a week is bleeding $45,000 to $120,000 a year in booked revenue that simply walked to a competitor. That is not soft pipeline. That is a customer who had their wallet out and dialed someone else.
The reflex answer is voicemail. Voicemail does not save the call. Around 85% of callers who reach a voicemail box do not call back, and roughly 80% of those calls leave no message at all. The caller is not leaving you a to-do item. They are gone. We unpacked the full arithmetic in what a missed call costs your business, and the short version is brutal: the leak is bigger than most owners believe because it is invisible. You never see the calls you did not answer.
Why hiring a front desk does not fix it
The obvious move is to hire someone to answer phones. The numbers make that a hard trade. The U.S. Bureau of Labor Statistics reported a median receptionist wage of $17.90 per hour, about $37,232 per year, in May 2024, per the BLS Occupational Outlook Handbook. That person works one shift. They go home at 5pm. They take lunch, get sick, and quit.
And the role itself is hard to staff. The same BLS data shows little to no employment growth for receptionists from 2024 to 2034, with around 128,500 openings projected each year, almost all of it replacement churn. People cycle through the seat constantly. You are not hiring once. You are hiring, training, and re-hiring on a loop, and paying recruiter and ramp costs each time.
Even a perfect hire covers maybe a third of the week. Nights, weekends, the 2am no-heat call in January, the Saturday flooding call - those still go to voicemail. You cannot afford three of them to cover every hour. So the structural gap stays open.
Voicemail vs human vs answering service vs AI receptionist
There are four real ways to handle the front door. Most operators have tried the first three and live with the gaps. Here is the honest comparison.
| Capability | Voicemail | Human receptionist | Answering service | AI receptionist (Front Door Loop) |
|---|---|---|---|---|
| Cost per month | ~$0 | ~$3,100 loaded | $200 - $1,500 | from EUR 1,800 |
| Hours covered | None (just records) | ~40/week, one shift | Often 24/7 | 24/7, every call |
| Books into your calendar | No | Yes | Rarely | Yes, into your scheduler |
| Dispatches emergencies | No | Yes, if on shift | Takes a message | Yes, warm-transfer to on-call |
| Qualifies job value | No | Yes | No | Yes, by trade rules |
| Scales with call spikes | n/a | No, one person | Partly, with hold times | Yes, parallel calls |
The answering service column is the one people misread. A traditional answering service is a message taker. It does not book the job. It hands you a callback list you still have to work once the truck is back in the yard, by which time the customer has called someone else. We drew that line in detail in AI voice agent vs answering service. The difference is not warmth. It is the calendar. A message is an open loop. A booking is a closed one.
The named system: luup's Front Door Loop
We build one thing for service-ops operators and we call it the Front Door Loop. It is not a chatbot bolted onto your number. It is a closed loop that turns a ring into a booked job without a human touch, and the philosophy behind it lives in the Front Door Loop and our broader Closed Loop Score framework.
Here is what happens on a call:
- Answer in under 90 seconds. Every call, every hour. That is the SLA. No ring-out, no voicemail, no hold music while a crew member decides whether to climb down.
- Qualify the job. The agent runs your trade's intake script: what is the problem, where, when, is this a quote or a booking. It captures the data your dispatcher needs.
- Book or dispatch into the scheduler. The appointment is written straight into your CRM or scheduling tool. Not a message. An actual slot.
- Text a confirmation. The caller gets an SMS with the time and what to expect, so the booking sticks and no-shows drop.
- Escalate true emergencies. Gas smell, no heat in a cold snap, active flooding - the agent screens for emergency intent first and warm-transfers or texts your on-call human immediately. Routine work flows into the calendar. The dangerous stuff jumps the queue.
We build it in 5 days. Pricing starts from EUR 1,800 per month. That is less than half the loaded cost of a single front-desk hire who covers one shift, and it covers all of them. The vertical build for trades lives at voice agents for home services, and the broader explainer is in AI voice agent for home services.
The stack we actually build on
No mystery box. We build on tools you can name and audit. For the voice layer we use Vapi or Retell as the orchestration layer, with ElevenLabs voices when we want a specific warmth on phone speakers. Telephony runs on Twilio. The dispatch and CRM wiring - the part that writes the booking into your scheduler and fires the SMS - runs on Make.com or n8n. If your scheduler is on a connector we support, it plugs in. If it is not, we wire it with a webhook.
The engines are not the product. The script logic, the routing rules, and the emergency-escalation tree are the product. That is the boring 80% that takes the 5 days, and it is the part that decides whether the agent books jobs or annoys callers. You can hear the difference yourself in the voice agent sandbox before you commit to a build.
Why the timing is real, not hype
Two things changed. The technology crossed the line where a phone call sounds like a person, and the buyers now expect it. Gartner reported that 85% of customer service leaders planned to explore or pilot conversational generative AI in 2025. This is not a fringe experiment anymore.
The trajectory is steeper. Gartner also predicts agentic AI will autonomously resolve 80% of common customer service issues by 2029, with a 30% cut in operational costs. The productivity gains are already measured, not theoretical: McKinsey found that generative-AI-assisted agents resolved 14% more issues per hour and cut handle time by 9%. The point for a trades operator is simple: the cost curve and the customer-acceptance curve both moved. Waiting another year just means another year of the leak.
The decision framework: run your own number
Skip the vibes. This is a math decision, and the model fits on one line:
(monthly missed calls x close rate x average job value) vs the AI receptionist's monthly cost.
Work an example. Say you miss 30 calls a month, conservative for a shop where the crew is on site all day. Say you would close 30% of them if you actually picked up. Say your average job is worth $400. That is 30 x 0.30 x $400 = $3,600 a month in recovered booked revenue. Against a Front Door Loop from EUR 1,800 a month, the agent pays for itself roughly twice over, before you count the 24/7 coverage and the no-show reduction from SMS confirmations.
Change the inputs to yours. A higher average ticket, like HVAC replacements or panel upgrades, moves the math hard in your favor fast. The model breaks even at surprisingly low call volume once the average job clears a few hundred dollars. Plug your real numbers into the Revenue Leak Heatmap and the Open Loop Tax Calculator. If the recovered revenue does not clear the monthly cost with margin, do not buy. That is the whole test.
The honest part: who this is NOT for
An AI receptionist is the wrong move for some shops, and we would rather tell you now than sell you a tool that disappoints. The system review we run is documented in our audit of 50 mid-market AI stacks, and the pattern is clear: tools fail when the fit was wrong from the start.
- Under 20 calls a month. The build cost will never pay back at that volume. A human or even voicemail with diligent callbacks is the cheaper honest answer. Come back when the phone rings more.
- Jobs that need deep human technical triage on the first call. If your first conversation is a senior tech diagnosing a complex system over the phone, an automated qualifier will get in the way. The agent is great at intake and booking, not at diagnosis.
- Dirty scheduler or CRM data. Garbage in, bad bookings out. If nobody owns clean records and consistent fields, the agent will write appointments into a mess. Fix the data discipline first, then automate. Automating chaos just produces faster chaos.
If none of those describe you, the math almost always works for a dispatch-driven trade. If one of them does, the responsible answer is to wait or to fix the prerequisite first. We say no to bad-fit shops because a tool that disappoints costs us more in reputation than the contract is worth.
How to find out in 10 minutes
You do not need a sales call to know whether this fits. You need two numbers: how many calls you miss, and what an average job is worth. Take the Closed Loop Score quiz to find your worst-leaking loop, then run the figures through the Revenue Leak Heatmap. If voice is your biggest leak and the math clears, the deeper proof is in our case studies, where the same loop closes for real operators. The full service spec sits on the voice agents page.
When you are ready to scope a build, tell us your call volume and your scheduler and we will tell you straight whether the Front Door Loop pays back for your shop or whether you are better off with a human for now. We ship in 5 days. The leak has been open longer than that already.
Frequently asked questions
What is an AI receptionist for service businesses?
An AI receptionist for service businesses is an inbound voice agent that answers every call in under 90 seconds, qualifies the caller, books the job into your scheduler or CRM, texts a confirmation, and escalates real emergencies to your on-call human. It runs 24/7 and does not go home at 5pm or take a lunch break while your crew is on site.
How much does an AI receptionist cost compared to hiring a front desk person?
luup builds the Front Door Loop in 5 days from EUR 1,800 per month. The U.S. Bureau of Labor Statistics put the median receptionist wage at $17.90 per hour, roughly $37,232 per year, and that person covers one shift, not nights or weekends. The AI receptionist covers every hour for less than half the loaded cost of a single hire.
Is an AI receptionist better than an answering service for contractors?
An answering service takes a message. An AI receptionist for service businesses books the job. The difference is the calendar. A message service hands you a callback list you still have to work after the truck is back. The AI receptionist qualifies job value, writes the appointment into your scheduler, and only escalates true emergencies, so the loop closes without a human touch.
Will an AI receptionist handle emergency calls correctly?
Yes, when the routing is built right. The Front Door Loop screens for emergency intent first - gas smell, no heat in winter, flooding - and warm-transfers or texts your on-call tech immediately. Routine bookings flow into the scheduler. The escalation rule is defined before launch, so a true emergency never sits in a booking queue.
Which service businesses should NOT buy an AI receptionist?
Businesses under 20 calls a month will never recoup the build cost. Trades whose first call needs deep human technical triage are a poor fit. And anyone unwilling to keep their scheduler and CRM data clean should fix that first, because garbage in means bad bookings out. For those operators, voicemail or a human is the cheaper honest answer.
The leak is the only number that matters. Run yours on the Revenue Leak Heatmap, take the Closed Loop Score quiz, and if voice is your worst loop, tell us your call volume. We will tell you honestly whether a Front Door Loop pays back for your shop.