The Real Cost of Customer Support for Small Online Stores
This is a cost model with stated assumptions, not a survey. All numbers below are computable from the inputs shown. Use the calculator to plug in your own figures.
Customer support is the cost small stores feel but rarely model. You know you're spending time on it — but how much, exactly, and would AI change the math? This paper gives you a transparent way to estimate it and compare the two paths: handling support yourself (or with a hire) versus pay-per-use AI that drafts and answers for you.
The model
We compare monthly support cost on two paths.
cost = emails × minutes_per_email × wage_per_minute + voice_minutes × voice_handling_cost + tooling
Human handling also carries hidden overhead: training, management, idle time when volume is low, and coverage gaps outside work hours.
cost = emails × $0.20 + voice_minutes × $0.10
No seats, no base fee, no idle cost. You pay only for what you actually send or answer.
Assumptions (stated so you can challenge them)
- Human handling time: 8 minutes per email (read, research policy, write, send) — a conservative mid-estimate for non-trivial support.
- Loaded wage: $18/hour ($0.30/minute) for support labour including overhead. Adjust in the calculator.
- Voice handling: human call handling valued at the same $0.30/min; AI voice at $0.10/min.
- Tooling: $30/month baseline for a shared inbox/help-desk on the human path.
- AI price: $0.20 per email reply, $0.10 per voice minute (RespondThat published rates).
Three volume scenarios
| Scenario | Emails / mo | Voice min / mo | Human cost / mo | AI cost / mo | AI saving |
|---|---|---|---|---|---|
| Quiet store | 200 | 60 | $84 | $46 | ~45% |
| Growing store | 800 | 200 | $326 | $180 | ~45% |
| Busy store | 2,500 | 600 | $1,010 | $560 | ~45% |
Human cost = emails×8×$0.30 + voice×$0.30 + $30 tooling. AI cost = emails×$0.20 + voice×$0.10. Percentages shift in AI's favour as volume rises because human cost is linear with overhead, while AI has no fixed base.
What the model shows
Key findings
- At every scenario above, pay-per-use AI is cheaper — and the gap widens with volume because there's no seat or base fee.
- The human path's real drag isn't the per-email rate; it's idle time and coverage gaps you still pay for on quiet days and after hours.
- The crossover where a dedicated hire beats AI is typically in the low thousands of contacts/month — below that, salary plus management rarely pays off.
- For spiky businesses (holiday peaks), per-use scaling to zero in slow periods is a structural advantage seat pricing can't match.
Live calculator
Recompute with your own numbers. All fields editable; the model updates instantly.
| Path | Monthly cost |
|---|---|
| Human (in-house/hired) | — |
| AI pay-per-use | — |
| You save | — |
Limitations
- This is a planning model, not a measured benchmark. Your handling time, wage, and mix of email vs voice will differ.
- It excludes quality differences — a great human reply and a good AI draft aren't identical, and a human review step is still recommended for nuanced cases.
- AI price uses RespondThat's published rates; other vendors differ. The model is vendor-agnostic except for that one input.
Frequently asked
Using a transparent model (human handling time ~8 minutes per email at $18/hour, plus tooling), 500 emails/month is roughly $120 in labour alone. The same volume handled by pay-per-use AI at $0.20 per reply is about $100 — and drops further as volume grows because there are no seats or base fees.
For most small stores with variable or modest volume, pay-per-use is cheaper because you pay only for replies sent or minutes answered. The crossover versus a dedicated hire is typically in the low thousands of contacts per month — below that, a salary plus management overhead rarely pays off.
No. The figures here are outputs of a stated cost model with explicit assumptions (handling time, wage, AI price), not survey data. They are meant as a planning tool you can recompute with your own numbers using the calculator on the page.