The 30-Day AI Receptionist Report Card: 7 Numbers That Tell You If It’s Working

Thirty days in, what do you actually know?

You’ve had an AI receptionist answering your phones for a month. The calls are being handled. The question from you, your partner, or your accountant is the obvious one: is it working?

“Working” is vague, and vague questions get vague answers. What you need is a report card: a short list of numbers you check at the 30-day mark that tell you, honestly, whether the system is earning its keep. Here are the seven that matter most.

1. Answer rate

What percentage of inbound calls were answered vs. going to voicemail or ringing out? This is your headline number. Before an AI receptionist, small businesses commonly answer somewhere between 38% and 72% of calls (411 Locals, 2016; CallRail, 2026). After, it should be at or near 100% for the hours the system covers. If it’s not, something is misconfigured: calls aren’t routing to the AI, or the AI’s schedule doesn’t match your actual hours.

2. Booking rate

Of the calls the AI answered, what percentage turned into booked appointments or jobs? This is the number that connects phone coverage to revenue. A healthy booking rate depends on your industry, but the trend matters more than the absolute value. Compare it to your best estimate of the before period. If booking rate holds steady while answer rate doubles, you’ve roughly doubled your booked jobs.

3. After-hours capture

How many calls came in outside your old business hours, and how many of them became bookings? This is pure found revenue: calls that previously went to voicemail, where 85% of callers never leave a message (Ruby Receptionists, 2025). Many businesses find that 20 to 30% of their AI-handled calls arrive after hours. Multiply those bookings by your average job value and you have the clearest ROI line on the report card.

4. Escalation rate

What percentage of calls did the AI hand off to a human? Some escalation is healthy: complex situations, upset callers, and judgment calls should reach a person. But if escalation is very high, the AI’s scripts may be too narrow. If it’s near zero, check whether callers with real problems are getting what they need or just getting processed. The right number is “enough that humans handle the exceptions, few enough that the AI earns its keep.”

5. Call-back and no-show rate on AI-booked appointments

Appointments the AI books should show up at roughly the same rate as appointments your staff books. If AI-booked appointments no-show more often, the confirmation workflow needs work: reminders, the right timing, the right channel. This confirmation workflow guide covers what actually reduces no-shows.

6. Caller sentiment

Numbers don’t capture everything. Listen to a sample of calls, maybe ten across the month, and ask a simple question: did the caller get what they needed, and did the interaction sound like your business? Our guide to reviewing your first 100 AI-answered calls gives you a routine for this. One awkward interaction pattern, repeated hundreds of times, is worse than one bad hire having a bad day.

7. Cost per booked job

Add up the month’s AI receptionist cost and divide by the number of jobs it booked. Compare that to your other acquisition channels: ads, lead platforms, referrals. For most businesses, the AI’s cost per booked job lands well below paid channels, because it’s converting demand that already exists rather than buying new demand. Here’s the full breakdown of the cost-per-booking metric if you want to go deeper.

What good looks like at 30 days

A healthy first month usually shows: answer rate near 100%, booking rate at or above your historical rate, meaningful after-hours capture, escalation in the low teens, and a cost per booked job that beats your paid channels. If you’re seeing that, the system is working. Optimize the scripts and keep going.

If you’re not seeing it, don’t scrap the system. Diagnose it. Low booking rate with high answer rate is a script problem. High escalation is a scope problem. Low after-hours capture might mean your callers don’t know you’re now answering after hours, which is a marketing problem, not a phone problem.

The 30-day report card isn’t a verdict. It’s a baseline for month two. Run the same seven numbers monthly, and you’ll watch the system get better as you tune it. That’s the real ROI story: not a single number, but a number that keeps improving.