Bolti AI - Automated Call Agent Platform Logo

Per-Minute Billing for Voice AI: What Actually Gets Charged

Bolti Team·

Bolti is a voice AI platform for building production-ready phone agents — and like most voice AI platforms, it charges at ₹7/min on a pay-as-you-go basis. Before you commit to any voice AI tool, you want to know precisely what triggers that meter. This post removes the ambiguity so you can budget accurately from day one. Start with 50 free minutes and test against real call data.

What is per-minute billing for voice AI?

Per-minute billing means you pay for the duration of each phone call your AI agent handles — typically rounded to the nearest second or 6-second block, starting from the moment the call connects until it ends. You are not charged for idle time, setup, or failed dial attempts that never connect.

The key word is connects. Most platforms, including Bolti, start the clock when the called party picks up — not when your system dials. So a call that rings for 20 seconds and goes to voicemail without connecting costs you nothing (or is handled by a separate voicemail-detection policy).

What the per-minute rate actually covers

A single ₹7/min line item bundles several underlying costs:

  • Telephony — the PSTN or SIP carrier leg that carries the audio
  • STT (Speech-to-Text) — transcribing the caller's voice in real time
  • LLM inference — the language model deciding what to say next
  • TTS (Text-to-Speech) — synthesizing the agent's reply into audio

Each of those layers has its own provider cost. Bolti's pipeline runs STT → LLM → TTS in a streaming loop many times per call, so the per-minute rate is a blended number across all four components — not just the phone line.

How much does voice AI cost per minute in 2026?

Rates across the market in 2026 range from roughly ₹5 to ₹25 per minute depending on the provider, the LLM model chosen, and the TTS voice quality. Bolti charges a flat ₹7/min pay-as-you-go with no seat fees or monthly minimums.

Here is a rough comparison of cost structures you will encounter:

Billing model How it works Watch out for
Flat per-minute One rate regardless of call complexity Make sure LLM + TTS are included
Component-based STT + LLM tokens + TTS billed separately Costs spike with long agent turns
Seat / subscription Monthly fee + per-minute overage Underuse wastes the base fee
Per-call flat Fixed fee per connected call Expensive for short calls

Bolti uses the flat per-minute model — you pay ₹7 for every minute of connected call time, and the STT, LLM, and TTS costs are included.

What exactly counts as a billable second?

Billable time begins the moment the call is answered and ends when the call is terminated by either party. Every second in between is counted — including:

  • Silence while the caller is thinking
  • Hold-like pauses between agent turns
  • The agent's TTS audio playing back
  • Time spent waiting for a tool call (e.g. a CRM lookup) to return

What is not billable:

  • Ring time before the call connects
  • Calls that fail to connect (busy, no answer, network error)
  • Time after the call ends but before your system processes the recording

Why silence still counts

The voice pipeline does not pause when neither party is speaking. Bolti's voice activity detection (VAD) and turn-detection are running continuously — listening for the caller to resume, handling potential interruptions, streaming the next LLM response. That compute runs whether or not audio is flowing, so the meter stays on.

Tool calls and API latency

If your agent calls an external tool — a calendar API for appointment booking, a CRM for customer data, a payment gateway — the time spent waiting for that API to respond is part of the connected call and is therefore billable. Keep your tool response times under 1–2 seconds to avoid inflating call duration unnecessarily.

How does provider choice affect the cost per call?

Bolti's flat ₹7/min rate holds regardless of which STT, LLM, or TTS provider you configure for that agent. However, provider choice directly affects call duration, which affects your total bill.

  • Faster STT (e.g. Deepgram or Fennec for Indian languages) means shorter perceived pauses, so callers stay engaged and calls resolve faster.
  • Faster LLM inference (e.g. Groq-backed models) reduces the gap between the caller finishing a sentence and the agent starting its reply — sub-second turn-taking keeps calls tight.
  • Concise TTS output — if your system prompt produces long-winded responses, the agent talks longer, the call runs longer, and you pay more.

In practice, a well-tuned agent on Bolti handles a typical appointment-booking call in 2–4 minutes. A poorly tuned one with slow tools and verbose prompts might run 7–8 minutes for the same task — more than doubling your cost per outcome.

See Bolti use cases for benchmarks across outbound sales, HR screening, payment reminders, and support.

How to estimate your monthly voice AI bill

Use this simple formula:

Monthly cost = (calls per day × average call duration in minutes × ₹7) × 30

Example: 200 calls/day × 3 min average × ₹7 × 30 days = ₹1,26,000/month

Steps to build a realistic estimate:

  1. Measure your current call handle time — pull the average from your existing call logs or CRM.
  2. Apply a 10–20% buffer — AI agents are often faster than human agents on structured tasks, but add buffer for edge cases.
  3. Multiply by ₹7 — no hidden fees, no per-seat charges.
  4. Add tool latency — if your CRM or booking system is slow, add 15–30 seconds per call to your average.

For high-volume operations (10,000+ calls/month), contact Bolti for volume pricing at bolti.co.in/contact.

Common billing questions answered

Does a 30-second call cost less than a 1-minute call?
Yes. At ₹7/min, a 30-second call costs ₹3.50. Billing is prorated per second, not rounded up to the full minute.

What if the caller hangs up immediately?
If the call connected (was answered), you are billed for the seconds it was live — even if that is only 5 seconds. If it never connected, you pay nothing.

Are inbound and outbound calls billed the same way?
Yes. The per-minute rate applies to both inbound calls (caller dials your agent's number) and outbound calls (your agent dials a contact). What matters is connected duration, not call direction.

Set up your first voice AI agent and see real costs

The fastest way to understand your actual per-minute costs is to run real calls. Bolti gives you 50 free minutes to do exactly that — enough for 15–25 typical calls across any use case. No credit card required to start. At ₹7/min pay-as-you-go after that, you can scale gradually without committing to a monthly plan. Start your free trial and build your first agent in under 10 minutes.

Frequently Asked Questions

How much does voice AI cost per minute in India?

Rates vary by provider and configuration. Bolti charges ₹7 per minute on a pay-as-you-go basis, with no monthly minimums or seat fees. This rate covers telephony, speech-to-text, LLM inference, and text-to-speech in a single blended charge.

What counts as a billable minute with voice AI?

Billing starts when the call connects (the called party answers) and ends when the call terminates. Every second in between is counted — including silence, tool-call wait time, and TTS playback. Ring time before connection and calls that never connect are not billed.

Does call duration include time waiting for API or tool responses?

Yes. If your voice AI agent calls an external API — for CRM lookups, appointment booking, or payment checks — the time spent waiting for that response is part of the connected call and is billed at the per-minute rate. Keeping tool response times under 1–2 seconds helps control costs.

Is per-second billing the same as per-minute billing?

Most per-minute billing is actually per-second billing prorated to a per-minute rate. So a 45-second call costs 75% of the per-minute rate, not a full minute. Always confirm whether your provider rounds up to the full minute or prorates per second.

How many free minutes does Bolti offer for testing?

Bolti gives new accounts 50 free minutes with no credit card required. That is enough for 15–25 typical phone calls across use cases like appointment booking, outbound follow-up, or HR screening.