Invoca's 2026 home services benchmark report puts the share of calls answered by a person at 52%. Filter out the misdials by looking only at calls lasting more than thirty seconds and it improves to 73%, which still means roughly one in four people who genuinely tried to reach a home services business did not reach anyone. Across sub-industries the answer rate ranges from 32% to 74%.
That is the entire market for answering services in one statistic. The question is which kind to buy, and whether you need one at all.
Three models, not one
The category looks like a single product and behaves like three.
Human answering services put a trained person on every call. Ruby is the reference point here, over twenty years in business with US-based receptionists, and it bills by the minute. Entry pricing depends on which source you read, landing somewhere between about $235 and $320 a month for 50 to 100 included minutes, with per-minute overage in the range of $4.49 to $6.39. Higher tiers climb steeply, with reports of $650 to $720 a month at 200 minutes.
Hybrid services route routine calls to AI and escalate the rest to a person. Smith.ai works this way and bills per call rather than per minute, which changes the risk profile: an eight minute call costs the same as a one minute call. Its AI tier starts around $97.50 a month for a capped number of calls, with human-handled plans starting near $292.50. Spam calls do not count against the plan, which matters more than it sounds like it should.
AI-only services are flat rate and much cheaper. This tier runs roughly $25 to $199 a month, with several providers offering unlimited inbound calls at the top of that range rather than metered minutes.
A caveat on all of these numbers: nearly every published comparison in this category is written by a competitor. The figures above are drawn from several such sources and they disagree with each other, particularly on Ruby. Treat them as the right order of magnitude and get a current quote before you decide anything.
The pricing trap is the meter
The structural risk in this category is not the monthly price. It is per-minute billing colliding with your busy season.
A plan with 50 included minutes covers about seventeen three-minute calls. In a normal month that might be fine. In the week the first hard freeze hits, or the first ninety degree day, your call volume triples and every minute past the cap bills at four to six dollars. The bill that read $245 in March reads considerably more in July, and it does so precisely when you are least able to sit down and review it.
Flat-rate and per-call models remove that variance. That is their real advantage, more than the sticker price.
What AI actually does well
Honest assessment: for scheduling, taking a message, answering hours and service area questions, and basic intake, current AI voice handling is good enough that most callers will not particularly notice or care. Estimates put that at something like 60 to 80% of typical business calls.
The remaining 20 to 40% is where it matters. A caller with water coming through the ceiling, an elderly customer who is confused and frightened, a commercial account with a complicated multi-site question. Those calls need judgment, and a caller who feels handled by a machine during an emergency remembers it.
So the real question is not whether AI is good enough. It is what fraction of your calls are the hard kind, and whether the service you pick escalates properly when it hits one.
Versus hiring someone
The comparison people reach for is a full-time receptionist. Bureau of Labor Statistics data puts the median receptionist wage around $37,230 a year, and loaded cost with benefits and payroll taxes typically lands above $45,000.
Almost no local service business with one crew has enough call volume to justify that, and the person would be idle most of the day. The more honest comparison is against what you are doing now, which is usually answering the phone yourself from a job site, or not answering it.
The part nobody selling this will tell you
An answering service converts a missed call into a captured message. It does not convert a captured message into a booked job. If nobody calls those messages back within the hour, you have paid a monthly fee to build a more organized list of people who hired someone else.
That is the actual failure mode. The service works exactly as advertised, the messages arrive, and they sit unreturned until the customer has already booked with whoever picked up.
The honest read
Work out roughly how many calls you miss in a week and what a booked job is worth. If you are missing five calls a week and your average job is $400, the arithmetic makes almost any option in this category look cheap, and a flat-rate AI service at $50 to $199 a month is a reasonable place to start.
If you are missing one call a month, this is a solution looking for a problem, and the money belongs somewhere else.
And if you buy one, decide before you sign who calls the messages back and how fast. That decision determines whether this is an investment or a subscription.