The conversation about AI in small business is mostly wrong. It is framed as "AI vs no AI," which is a software decision, and software decisions are commodity decisions. The framing should be: AI vs the headcount you cannot afford to hire. That is a labour decision, and labour decisions are valued differently because the alternative is not "do nothing" — the alternative is "lose the revenue because you cannot staff the function."
When you frame the buy that way, the math is obscene. And the math is what closes deals.
| Vertical | Role being replaced | Loaded annual cost | AI replacement cost | Savings per year |
|---|---|---|---|---|
| Dental | Front-desk receptionist | $55–72K | $300–1,200/mo ($3.6–14.4K) | $40–68K |
| Dental | Second front-desk hire | $52–72K | $300–1,200/mo | $40–65K |
| Dental | Insurance verifier | $42–55K | $250–500/mo | $36–49K |
| Med spa | Patient coordinator | $48–65K | $250–650/mo | $40–57K |
| HVAC | Dispatcher (1 per 8–10 techs) | $55–72K | $350–650/mo | $47–64K |
| HVAC | CSR / call taker | $38–52K | $350–650/mo | $30–44K |
| Roofing | Supplement writer | $60–90K | $500–1,000/mo | $48–78K |
| PI law | Intake paralegal | $55–75K | $500–1,000/mo | $43–63K |
| Plastic surgery | Patient concierge | $50–70K | $500–1,000/mo | $38–58K |
These are 2026 numbers, sourced from independent vendor pricing, BLS wage data, and observed fully-loaded employer cost ratios in the 1.25–1.45× band. They will drift. The shape of the curve will not. An AI agent does the work of a person at 5–10% of the cost, and it does not call in sick.
A solo dental practice with two ops, ~180 inbound calls/week, and one front-desk hire:
Read that again. A $400–800/month platform fee returns $50K–$100K in recovered revenue and avoided labour cost. That is a 30–100× return on the monthly subscription.
Two reasons. First, they anchor to the wrong comparison. They compare $400/month for an AI tool to $0 (do nothing) instead of $4,000/month (the next hire). Second, they buy the wrong thing. They buy a chatbot that takes messages and emails them to the team. That is a $50/month product with a $300/month wrapper. The product that delivers the ROI is the one that writes the appointment back to the practice management system in real time — checks the live schedule, books the slot, attaches the intake notes, sends the confirmation, updates the chart.
Read-only is voicemail with better hold music.
Different call volumes need different deployments. The rule of thumb from operator data (Q1 2026):
Most practices and most service SMBs sit in the bottom two bands. For them, AI alone closes the gap.
A useful number for anyone thinking about building, investing in, or partnering on AI agent services:
The structural reason these margins exist: the customer is not paying for the compute. The customer is paying for the outcome (more appointments booked, books closed, leads worked). The compute is incidental. The framing is what creates the margin.
The question is not "should I buy an AI tool?" The question is: "Should I pay $1,500/month for an agent that recovers $50K/year in lost revenue and replaces a $55K hire I cannot afford to make?" That question has one answer. The only variable is which vendor you trust to do it without creating a compliance problem in the process.
Pick the vendor whose AI writes back to the system of record, logs every action, signs a BAA, and hands you an audit trail on demand. That vendor is the one you can defend in front of a regulator, a banker, or your own board.
CEO