Quick answer (updated August 2026): Measure five things before buying an AI receptionist: missed calls, the share that were real job inquiries, your close rate on answered inquiries, your average ticket, and how many important calls happen after hours. If those numbers are weak, the tool is not the fix. If they are strong, run a 30-day trial with a real scoreboard, not a demo-day impression.

This is a measurement problem before it is a software problem

Most AI receptionist pages jump from “missed calls are bad” straight to vendor features. That skips the only part that decides whether the tool is worth it in your business: how much revenue the missed-call leak actually puts at risk, and whether the call handoff after the answer is strong enough to convert that answer into a booked job.

The evidence says the leak is real. Invoca’s 2024 call-tracking data found 27% of calls to home-services businesses go unanswered, and fewer than 3% of callers pushed to voicemail leave a message. CallRail’s 2025 analysis of 1.1 million business calls put the home-services miss rate lower, at 14%. The range matters less than the method: use your own logs if you have them, and use 14% as the floor if you do not.

Hourback’s view is simple: measure first, buy second. We sell no phone software and take no commissions, so the worksheet below is meant to help you not buy the tool when the numbers do not support it.

The five numbers that matter

You can gather the first four in about 15 minutes from a phone log, dispatch board, and a rough average-ticket number from your books.

Number What to count Why it matters
Missed calls per week Calls that rang out or hit voicemail This is the size of the leak
Job-inquiry share The share of missed calls that were real work, not spam or vendors Not every call was revenue
Close rate on answered inquiries What share of answered job inquiries become booked work This turns leads into dollars
Average ticket Average revenue from the jobs that do close This sets the value of each recovered call
After-hours or overflow share How many important calls came when nobody was available This tells you whether 24/7 coverage is the real benefit

One more number belongs on the trial scoreboard, not the pre-purchase worksheet: handoff failure rate. Count wrong bookings, duplicate bookings, bad escalations, or callbacks your staff had to fix. A tool that answers every call and creates dispatch chaos is not working.

The 15-minute worksheet

Use one normal business week. Round down on value, and round up on tool cost.

Step 1: Count the misses

Pull a week of inbound calls from your phone system. Count how many were not answered by a person.

  • If you have the exact number, use it.
  • If you do not, estimate using your weekly volume times a miss-rate range.
  • Use Invoca’s 27% as the aggressive case and CallRail’s 14% as the floor.

Step 2: Remove the junk

Not every missed call was a paying job. Back out spam, wrong numbers, recruiters, and vendor calls. What remains is your job-inquiry share.

If you do not know, use 50% as a conservative haircut and revisit it after the trial.

Step 3: Use your real close rate

What share of answered job inquiries become booked work? Use your own number. If you do not track it, use a conservative range and note that uncertainty.

Step 4: Use your real average ticket

Average the actual revenue from completed jobs in the relevant category. Do not use the biggest invoice from the month. Do not use a vendor’s example.

Step 5: Run the formula

Weekly revenue at risk = missed calls × job-inquiry share × close rate × average ticket

Here is a conservative illustration, not a promise and not a client result:

Input Value
Weekly inbound calls 60
Missed-call rate 14% floor
Missed calls 8
Job-inquiry share 50%
Real missed inquiries 4
Close rate on answered inquiries 25%
Jobs lost 1
Average ticket $300
Weekly revenue at risk $300

At that floor case, you are looking at about $1,200 a month at risk against a tool that may cost $49 to $399 a month at published July 2026 prices. If your actual miss rate is closer to Invoca’s 27%, the number rises fast.

The Hourback decision rule

This is the part most vendor pages skip. A tool is worth a trial only if it clears all three tests:

  1. The leak is real. Your weekly revenue-at-risk number is clearly above the monthly tool cost.
  2. The calls matter. Enough important calls happen after hours, at lunch, or during overflow that a human cannot realistically catch them.
  3. The handoff can be designed. You can state, in writing, when the tool should book, when it should take a message, and when it should escalate to a person.

If any one of those fails, the answer is “not yet.”

That makes some honest “no” cases easy:

  • Low-value calls with a low average ticket
  • Very low missed-call volume
  • A business where every important call needs human judgment immediately
  • A broken booking or dispatch process that would only be automated into a larger mess

Exhaust the platform you already pay for first

This is the strongest objection to buying a new AI receptionist right now, and it is often correct. Many field-service and office platforms now ship some version of AI answering, missed-call text-back, or booking automation inside the suite you already use.

The suite-first rule is simple:

  • If your current platform already offers the feature, test that feature first.
  • If it does not, or if the feature does not handle your real call flow, compare standalone tools.
  • Do not buy a second system until you know the first one leaves a clear gap.

That is the same “existing tools first” rule now sitting in the decision queue for Hourback’s own prescriptions.

The 30-day trial scoreboard

If the worksheet says “yes, test it,” do not switch your number and hope. Run a 30-day measured trial.

Track these six numbers before and after:

Metric Before trial After 30 days
Calls answered
Calls missed
Jobs booked from phone leads
After-hours jobs captured
Wrong or messy handoffs
Monthly tool cost

Set the stop condition before launch:

  • Keep if answered calls rise, booked work rises, and handoff mistakes stay manageable.
  • Fix if answered calls rise but booking or handoff quality is weak.
  • Cancel if the tool mostly creates cleanup work for your staff.

The most useful question after 30 days is not “did callers like it?” It is “did the business keep more of the work it was already earning?”

When a paid assessment is the better first move

Skip the tool trial and buy a diagnosis first when the phone is only one leak among several. If calls, scheduling, quote follow-up, customer updates, and invoice chasing are all bleeding at once, the real problem is sequencing, not vendor selection.

That is where an AI opportunity assessment fits. It maps the phone against the rest of the week, applies the same measurement discipline, and tells you whether the first move should be an answering layer, quote follow-up, payment reminders, or no new tool at all. If we can’t save your business at least ten hours a week, your assessment is free.

If you already know the phone is the issue and only need the vendor comparison, start with our guide to AI receptionists for home services.

Bottom line

Measure before you demo. Count missed calls, remove the junk, use your real close rate and average ticket, and decide from the math whether the leak is big enough to justify the tool. Then run a 30-day trial with a handoff-quality scoreboard. The AI receptionist is worth testing when the missed-call leak is real, the after-hours demand is real, and the handoff rules are clear. Otherwise, fix the process first.

Sources