Instrument 11 / voice economics

Stop comparing voice agents by price per minute.

Model the loaded cost of every AI-resolved call after the voice stack, fixed fees, failed resolutions, and human handoffs.

Model one production month

Price the outcome, not the demo.

The defaults are an illustrative stack, not a vendor quote. Nothing is submitted or stored; every calculation runs in this browser.

Traffic and stack

Calls per monthConnected calls that incur usage cost. calls Average call lengthConnected minutes across successful and failed calls. min Platform / orchestrationAgent platform or realtime orchestration rate. $ TelephonyInbound, outbound, SIP, or WebRTC transport per minute. $ Speech to textStreaming transcription converted to a per-minute rate. $ Text to speechConvert character or token billing to talk-minute cost. $ LLM usageAverage model cost per connected conversation minute. $ Fixed monthly costSubscriptions, numbers, support, and reserved capacity. $

Outcome and handoff

AI resolution rateCalls completed correctly without a human handoff. % Human time per unresolved callEscalation, cleanup, and repeat-contact work. min Loaded human hourly costWages plus benefits, management, tools, and overhead. $ Current human cost per resolved callYour all-in baseline for the same call type. $

Read the economics

Every minute has layers.
Platform, transport, speech, and model costs compound.

Failure hits twice.
You paid for the call and still need a person to finish it.

Price the outcome.
A cheap minute can produce an expensive resolution.

Transparent method

What this voice AI cost calculator computes.

Voice stack cost adds the user-supplied platform, telephony, STT, TTS, and LLM rates. Monthly AI cost is that rate multiplied by connected minutes, plus fixed monthly fees.

Failure burden is unresolved calls multiplied by human cleanup time and loaded hourly cost. This makes handoffs visible instead of treating them as free.

Loaded cost per AI-resolved call divides total AI and handoff spend by calls resolved without a human. It is deliberately stricter than cost per minute or cost per attempted call.

Break-even resolution solves for the AI resolution rate at which loaded cost per AI-resolved call equals the stated human baseline. Validate resolution correctness, repeat contacts, abandonment, and customer outcomes before treating containment as success.

Voice economics guide

Use invoices and measured outcomes, not a vendor headline.

Rates change by provider, plan, region, direction, model, and volume. Replace the example values with current quotes or invoices, and use a verified resolution rate for the same call type.

Read the complete cost-per-minute formula → Open the AI receptionist worksheet → Model the latency of the same voice stack → Read ResearchAudio's voice AI architecture guide → Check current Twilio Voice pricing → Check current ElevenLabs pricing →

How do you calculate AI voice agent cost?

Add platform, telephony, speech-to-text, text-to-speech, and model cost per minute. Multiply by connected minutes, add fixed fees, then add the human work created by unresolved calls.

What is cost per resolved call?

This calculator divides all voice AI and handoff spend by calls the AI resolves without a person. That exposes paid attempts that fail, transfer, or create cleanup work.

Why does resolution rate change voice AI economics?

A lower resolution rate means fewer successful outcomes from the same usage bill and more human escalations. The denominator shrinks while the failure burden grows.

Which fees belong in a voice AI cost model?

Include orchestration, transport, STT, LLM, TTS, fixed subscriptions, phone numbers, capacity, support, and human escalation. Use current invoices or provider rate cards because pricing changes.