AI Receptionist ROI: The Math for a 5-Truck Shop

For a five-truck HVAC shop, the honest version of AI receptionist ROI is this: the plan pays for itself with one or two recovered jobs a month, and everything else is a measurement problem. The subscription price is small and knowable. The variable that decides the outcome is how many of your missed callers actually book when somebody finally answers, and almost nobody measures that before buying.

This article walks a full worked example with every assumption labeled as illustrative, shows what changes the answer, and lists the ways this kind of math gets quietly fudged by vendors. Where a number comes from a source, it is linked. Where it is an assumption, it says so.

Key takeaways

• The BLS median wage for receptionists was $18.27 per hour in May 2025, and benefits add about 30 percent on top of total compensation, which puts a fully loaded front-desk hour near $26.
• A 411 Locals monitoring study of 85 small businesses found 62 percent of calls went unanswered, though it is older data and describes small businesses broadly, not HVAC shops specifically.
• HomeGuide puts the average HVAC repair at $293, which is the ticket size the example below uses.
• Flat AI plans are priced between $79 and $249 per month with unlimited minutes (Goodcall), and AI retainers with CRM booking run $397 to $797 plus a $497 setup (AnswerForce).
• Illustrative example: for a shop missing 90 calls a month with a 25 percent booking rate on recovered calls, monthly gross profit from those jobs is about $3,000 against a plan cost under $900, even in the first month.
• The three assumptions that swing the result most are your missed-call rate, the recovery booking rate, and your gross margin. Change them honestly before believing any ROI number, including this one.

The five inputs you need

Any credible ROI calculation reduces to five numbers, and four of them are already in your systems.

Notice what is not on the list: the vendor's pricing page. Cost matters, but it is the least variable part of this equation. The AI receptionist pricing guide covers how the plans are structured once you know your inputs.

The worked example, labeled illustrative

This is an illustrative example, not a Praktivo client result, and not a forecast. It uses a five-truck HVAC shop with these assumed numbers.

The calls. The shop receives 450 inbound calls a month. Twenty percent, or 90 calls, currently end in voicemail, a hang-up, or nothing. That 20 percent is deliberately conservative: the 411 Locals study found 62 percent, and most busy shops we look at sit somewhere between. If your own missed rate is 10 percent, halve everything below.

The recovery. The AI receptionist answers all 90 that would have been missed. One in four of those callers books a job once a human or agent actually engages them. That is about 22 booked jobs. Is 25 percent realistic? It is a planning input, not a guarantee, and it is the single number worth testing hardest, because a dispatching desk that converts at 10 percent and a script that converts at 40 percent describe two very different businesses.

The value. At the HomeGuide national average of $293 per repair, 22 jobs produce roughly $6,600 in revenue. At an illustrative 45 percent gross margin, that is about $3,000 in gross profit, which is the number that should carry the decision. If you want the repair-cost context behind that ticket, our HVAC answering cost analysis breaks down the ranges.

The cost. Four published options, for comparison:

The verdict in the illustration. Against the $249 flat plan, recovered gross profit of about $3,000 produces a net of roughly $2,750 in month one, and one $293 repair covers the plan with money left over. Against the project build, the $5,000 build plus $300 management pays back inside two months at this run rate and then keeps contributing. Against the human per-call service, the same recovered work covers the bill with room to spare. None of these comparisons require the optimistic case to work; they require roughly one recovered job per month.

What changes the answer

Run the sensitivity yourself with your numbers, and treat this table as a demonstration of how much the booking rate moves the outcome.

Two observations. First, even the pessimistic case covers the flat plan and the retainer, which is why the subscription price is rarely the deciding factor. Second, the difference between 10 percent and 40 percent is a script and routing question, not a pricing question. The FAQ answers, booking rules, and escalation design described in our AI receptionist service are what move conversion, and they are also what make the optimistic column plausible or impossible.

If you want a lighter version of this same math before committing to anything, the missed-call revenue calculator walks through the per-call version.

The competing costs, with real numbers

A part-time human. BLS reports a median receptionist wage of $18.27 per hour, or $38,010 per year, with employment projected to decline 2 percent through 2035 as organizations automate administrative functions. Benefits matter: BLS counts benefits as 30 percent of total compensation for private industry workers, so dividing the wage by 0.70 gives a fully loaded rate near $26 per hour. A 20-hour-a-week role costs roughly $2,200 a month, works 20 hours a week, and can handle far more than phone calls. This is not an argument against hiring; it is an argument against comparing a 168-hour AI line to a fraction of a person.

A live answering service. Smith.ai's human plans are $300 per month for 30 calls, $810 for 90, and $2,100 for 300, with overage from $11.50 down to $8.50 per call. At the illustrative 90 recovered calls, the Starter tier plus overage lands near $990. The service is excellent and entirely human, and per-call pricing rewards answering rather than call length. The trade is price at volume and the depth of integration with your dispatch software.

Doing nothing. The cost of the status quo is not zero; it is the missed calls themselves, plus the owner hours spent returning them. The comparison that matters is not AI versus perfect coverage. It is AI versus what you have today.

What the math hides

The example above is honest about its assumptions but still simplified. Watch for these when you run your own version.

• Revenue is not profit. A recovered $293 repair with a 45 percent gross margin contributes $132, not $293. Vendors who stop at revenue overstate returns several times over.
• Not every missed call is a job. Some are spam, wrong numbers, vendors, and price shoppers who were never going to book. Assume a haircut, and check your call recordings to size it.
• Do not double count. If your office already answers 85 percent of calls well, the AI only owns the other 15 percent plus after-hours. The after-hours share is often where the real premium tickets live.
• Setup and management are real costs. A $497 setup or a $5,000 build belongs in the first-year math, as does any monthly management fee.
• Seasonality cuts both ways. A summer peak inflates missed-call counts and recovered revenue; January does not look like July. Judge on a full season, not a good week.
• The agent does not fix your offer. If your pricing is not competitive or your booking process is slow, recovered calls convert at the pessimistic end. AI buys coverage, not demand.
• Attribution is imperfect. The cleanest measurement is before-and-after booked jobs at similar ad spend, not a dashboard that claims revenue on a call that was never going to book anywhere else.

How to measure your own 30 days

1. Baseline week zero. Export 30 days: total inbound calls, calls answered by a human, calls that ended in voicemail or abandonment, after-hours share, and bookings recorded in your calendar or field service software.
2. Pick the comparison honestly. Keep ad spend and season roughly stable, or note the difference. If you install the agent during your busiest month, say so when you report results.
3. Track four numbers weekly after launch: calls answered, bookings created from calls, escalations completed correctly, and spam excluded from the counts.
4. Convert to gross profit monthly. Multiply recovered bookings by average ticket and margin, then subtract plan cost, setup amortized over 12 months, and any management fee.
5. Decide at day 30. If recovered gross profit does not exceed total monthly cost, narrow the scope to after-hours only and rerun. If it exceeds cost several times over, expand the agent to the main line and consider the next workflow. The five-truck shop in the illustration is not special; the arithmetic is just usually left undone.

Next step

Run step one this week: pull 30 days of call data and fill in the five inputs. That single exercise tells you whether this decision is worth more of your time. The six-step plan on our homepage turns the inputs into a priority order: start your AI automation plan. If your season is ending and you have a month of records, book a call and we will work the math with you. The HVAC case study shows what the configured system looks like in the field.

FAQ

What is a realistic ROI for an AI receptionist?

It depends on missed-call volume and average ticket, not the subscription price. In the illustrative five-truck example in this article, a flat AI plan pays for itself with one recovered repair per month. The uncertain variable is how many missed callers book when somebody finally answers, so measure that before and after.

How do I calculate AI receptionist ROI for my business?

Use five inputs: inbound calls per month, the share that go unanswered, the share of answered callers who book, average job value, and gross margin. Revenue math without margin overstates returns, and so does counting calls your team already answers.

Is an AI receptionist cheaper than hiring a receptionist?

On cash cost, usually yes. BLS puts the receptionist median wage at $18.27 per hour as of May 2025, and benefits account for about 30 percent of total compensation, so a 20-hour-a-week role costs roughly $2,200 a month fully loaded. The AI also answers at 2 am, while the human does far more than answer phones.

What assumptions make AI receptionist ROI math misleading?

Assuming every missed call would have booked, using revenue instead of gross profit, ignoring junk and spam calls, forgetting setup and management fees, and ignoring seasonality. Any single one of those can turn a break-even project into a fake windfall.