Most "AI ROI" posts are vibes. This one is arithmetic. We take 100 after-hours calls a month, run them through three real business types, and compare three options: a $150/month AI receptionist, a $1,200/month human answering service, and doing nothing. The answer is less obvious than the vendors want you to believe.
The inputs: four numbers that decide everything
Every receptionist ROI calculation reduces to four inputs:
- Call volume: 100 after-hours calls/month in all three scenarios.
- Answer rate: ~95% for AI or human services, ~20% for voicemail (most callers hang up).
- Close rate: the share of answered calls that become paying jobs — we use 25% for trades, 15% for law, 35% for dental.
- Average job value: $450 (HVAC), $2,500 (law firm intake), $800 (dental new patient).
We also assume 30% of calls are spam or wrong numbers — free to filter on AI plans, but billed minutes on human plans.
Scenario A: HVAC contractor ($450 avg job)
Of 100 calls, 70 are real. The AI answers 95% (67 calls) and the business closes 25% — ~17 jobs, or $7,650 in revenue against a $150 subscription. Doing nothing: voicemail catches 20% of the 70 real calls (14), closes 25% — ~3.5 jobs, $1,575. The AI's incremental value: roughly $6,075/month for $150. Payback period: about one day.
The $1,200 human service answers the same 95%, maybe closes slightly better at 28% thanks to the human touch — ~18.5 jobs, $8,325. Incremental over nothing: $6,750 for $1,200. It pays, but the AI captures 90% of the value at 12% of the cost.
Scenario B: law firm ($2,500 avg case)
Higher stakes, lower close rate. Of 100 calls, 70 real; AI answers 67, closes 15% — ~10 cases, $25,000 in revenue for $150. Doing nothing: 14 answered, 15% close — ~2 cases, $5,250. Incremental AI value: ~$19,750/month. The subscription is a rounding error.
Here's where the human service earns a hearing: a live receptionist might close 20% instead of 15% on sensitive intake calls — 13 cases vs. 10, or $7,500 more revenue for an extra $1,050 in cost. If your intake is emotionally charged (family law, criminal defense, personal injury), the human premium can genuinely pay. For transactional practices, it can't.
Scenario C: dental practice ($800 avg patient value)
Of 100 calls, 70 real; AI answers 67, closes 35% on appointment bookings — ~23 patients, $18,400 for $150. Doing nothing: 14 answered, 35% — ~5 patients, $4,000. Incremental: ~$14,400/month.
Dental is the cleanest AI win of the three: callers want one thing (an appointment), the script is simple, and the AI books directly into the schedule. Human services add almost nothing here — the decision is AI vs. voicemail, and it's not close.
The breakeven table
How many saved calls per month does each option need to break even?
| Business type | AI ($150/mo) | Human ($1,200/mo) |
|---|---|---|
| HVAC ($450/job, 25% close) | 1.4 calls | 10.7 calls |
| Law firm ($2,500/case, 15% close) | 0.4 calls | 3.2 calls |
| Dental ($800/patient, 35% close) | 0.6 calls | 4.3 calls |
Read that first column again: the AI receptionist breaks even if it saves one to two calls a month. At 100 calls, you're not deciding whether it pays — you're deciding how many multiples of payback you want.
The hidden costs most ROI posts ignore
- Spam and wrong numbers: ~30% of after-hours volume. Free on AI plans with screening; on per-minute human plans, 30 junk calls × 3 minutes × $5/minute = $450/month burned on nothing.
- Staff interruption: every call that reaches the owner's cell at 9pm has a cost beyond the minute — context-switching, family time, burnout. The AI absorbs 100% of this; voicemail absorbs 0%.
- Speed to callback: AI summaries land in your inbox in 2 minutes. Voicemail gets checked "in the morning," which in practice means 10am. In competitive trades, the 8-hour gap is where jobs die.
- Setup and script maintenance: budget 2–4 hours upfront and 30 minutes a quarter to update scripts for seasonal offers and price changes. Negligible, but not zero.
Want the worksheet?
Plug in your own call volume, close rate, and job value with our free True Cost of a Missed Call worksheet — it runs the same four-input math above in about 3 minutes. Grab it from the StackSmarter newsletter signup on our homepage.
Ready to pick a service? Our head-to-head comparison of Smith.ai, Ruby, and Quo Sona has the pricing tables and verdicts by business type.
FAQ
Isn't this math too optimistic? Not every call is a buyer.
That's why we haircut 30% for spam and use conservative close rates (15–35%). Even if you halve our close rates, the AI still breaks even at 1–3 saved calls a month. The math is robust because the subscription is cheap, not because the assumptions are rosy.
What if I only get 20 after-hours calls a month?
At 20 calls, the breakeven is still ~1–2 saved calls — but look at the budget tier. A $25–40/month call pack (Quo Sona-style) fits better than a $150 plan. Match the plan to the volume; don't buy 150-call capacity for 20 calls.
Does the human service ever beat AI on ROI?
Yes, in one case: high-value, emotionally charged intake (law, medical, high-ticket sales) where a human closes meaningfully more calls. The premium is ~$1,050/month over AI — it needs to save roughly one extra $2,500 case every 2–3 months to justify itself. Often it does.
What's the payback period, realistically?
For the AI option at 100 calls/month: days, not months. The honest caveat is month one — script tuning and test calls mean week one underperforms. Judge it at day 30, not day 7.