- Voice agents have two costs: a one-time build and ongoing per-minute usage.
- Usage stacks the voice platform, speech-to-text, the language model, text-to-speech and telephony.
- All-in usage commonly lands in the range of roughly $0.10-$0.30+ per minute, depending on choices.
- Focused builds typically cost a few thousand dollars; complex multi-system agents run into five figures.
- ROI comes from recovered calls and leads, calculate it from your own missed-call and close-rate data.
"How much does it cost?" is the first question every client asks about AI voice agents, and most answers online are either a vendor’s per-minute headline or a vague "it depends". Here’s the full picture: what you pay once, what you pay every month, and how to decide whether it’s worth it.
The two parts of voice-agent cost
- Build (one-time): conversation design, knowledge setup, integrations with your CRM and calendar, telephony, testing and a supervised pilot.
- Usage (ongoing): per-minute charges from the providers that make each call work, plus phone numbers.
What goes into per-minute usage
| Layer | What it does | What changes the cost |
|---|---|---|
| Voice platform | Orchestrates the real-time call (e.g. Vapi, Retell AI) | Platform plan and fees |
| Speech-to-text | Transcribes the caller | Provider and model |
| Language model | Decides what to say and do | Model size, prompt length, call length |
| Text-to-speech | Generates the agent’s voice | Standard vs premium voices |
| Telephony | Connects to the phone network | Country, inbound vs outbound, carrier |
Add them together and the all-in usage cost commonly lands somewhere around $0.10 to $0.30+ per minute. Premium voices, larger models, long prompts and international calls push it up; efficient configurations bring it down. Vendor pricing changes frequently, so always price your exact configuration.
Monthly running cost examples
| Scenario | Calls / month | Avg length | Minutes | At $0.15/min | At $0.25/min |
|---|---|---|---|---|---|
| Small office after-hours cover | 300 | 2 min | 600 | $90 | $150 |
| Busy clinic front desk | 1,500 | 3 min | 4,500 | $675 | $1,125 |
| Home-service company, all calls | 2,500 | 3.5 min | 8,750 | $1,313 | $2,188 |
| Outbound speed-to-lead | 4,000 | 1.5 min | 6,000 | $900 | $1,500 |
These are illustrative usage costs only, excluding phone-number fees, any platform subscription and optional support retainers.
What a build costs
Build cost depends on scope: how many call types, how many systems the agent must read from and write to, and how much testing and compliance work is needed.
| Build scope | Typical range | Examples |
|---|---|---|
| Focused agent | A few thousand dollars | After-hours receptionist, FAQ + booking into one calendar |
| Integrated agent | Mid four to low five figures | Qualification + CRM + calendar + SMS + transfers |
| Multi-agent system | Five figures+ | Several call types, custom APIs, squads, analytics |
Hidden costs to watch for
- Per-seat or per-agent licenses on some SaaS "AI receptionist" products.
- Premium voice upcharges that double usage cost for marginal quality gains.
- Long system prompts that inflate model costs on every turn.
- Compliance add-ons (e.g. HIPAA options) on certain plans.
- Tuning time, the first 30 days always surface edge cases; make sure it’s included.
How to calculate ROI
Voice agents pay back by recovering calls and leads you currently lose, and by freeing staff time.
Compare against monthly usage plus the build cost spread over 12 months.
Our Lead Loss Calculator estimates the first half of that formula in under a minute. For the broader method, see How to Calculate AI Automation ROI.
Buy vs build
Off-the-shelf AI receptionist apps are cheaper to start but limited to their templates and integrations. A custom build costs more upfront, integrates with exactly your stack and has no per-seat license. If your needs are standard, start with a product; if calls touch your CRM, scheduling rules and multiple systems, a custom agent usually wins on total cost within a year.
Written by Abdul Moeez, AI Automation Expert
Abdul designs and builds the AI voice agents, chatbots and automation systems Voxil AI ships: from conversation design and integrations to testing on real calls.