How Much Does an AI Voice Agent Cost per Minute in India?
Short answer: ₹2.00 to about ₹13 per connected minute for most setups. Indian platforms that route calls domestically sit at the low end; USD-billed global platforms at the high end once telephony and currency conversion are added. Here is how to work out your real number.
Updated
From ₹2.00/Min
Vistara AI costs ₹3.00/min pay-as-you-go with no monthly fee, or from ₹2.00/min on platform plans. No surge pricing or currency markups.
One Comparable Number
Ask every vendor for an all-inclusive monthly figure in rupees at your projected volume. The ones that resist are usually the ones whose headline rate excludes the most.
No Surge or Peak Pricing
The rate does not change because you are calling at a busy hour or during a festive period.
An AI voice agent in India costs between ₹2.00 and ₹13 per connected minute. Where a vendor lands in that range depends mostly on two things: whether calls are routed over Indian telephony, and whether you are billed in rupees or in dollars. The language model itself is a smaller share of the cost than most buyers expect.
Comparing voice AI pricing across vendors is harder than it should be, and not accidentally. Rates are quoted in different currencies, on different billing units, with different things included, against different volume assumptions. Two quotes that look thirty percent apart can be identical, or five times apart, depending on what each one leaves out.
This page gives you a method rather than a claim. Work through it for every vendor on your shortlist, including us, and you will end up with a single comparable rupee figure per month — which is the only basis on which a pricing decision can actually be made.
All-In Price per Minute: What Each Platform's Rate Includes
Headline per-minute prices are not comparable until you know what they include. The table lists each platform's published rate and whether telephony, speech recognition, the AI model and the voice are inside it. Prices were copied from each vendor's own pricing page on 13 Oct 2026; US-dollar prices are converted at about ₹92 per dollar.
| Platform | Published rate | Telephony included? | Speech, AI model and voice included? | Monthly fee |
|---|---|---|---|---|
| Vistara AI | ₹2.00–₹3.00/min | Yes (Jio, Airtel, Vi routes) | Yes | ₹0 (pay-as-you-go), ₹2,500 or ₹3,499 (platform plans) |
| Ringg AI | ₹6/min | Yes | Yes | Monthly minimum (amount not published); numbers ₹499/month |
| Bolna | $0.06/min (≈ ₹5.50) | No, added per minute (not published) | Yes, with bundled models; platform fee extra | None; $10 minimum top-up |
| Caller.Digital | ₹9.42–₹10.15/min | Yes | Growth plan only (bring your own keys on pay-as-you-go) | Growth needs 10,000 min/month |
| Bland AI | $0.12–$0.14/min (≈ ₹11–₹13) | No, billed by your carrier or Twilio | Yes | $0 (Start) or $299 (Build) |
| Vapi | $0.05/min + providers (≈ $0.08–$0.13 all-in) | Vapi numbers free; Twilio and others extra | No, each billed at provider cost | $0 to start; Core $29 |
| Retell AI | $0.07–$0.31/min | $0.015/min (SIP free) | Priced per component | $0 (pay-as-you-go) |
Gnani.ai, SquadStack, Exotel, Ozonetel and Verloop.io do not publish voicebot prices, so ask them for an all-in quote per connected minute. Full comparison of 12 platforms: best AI calling software in India.
A Comparison Method That Produces a Real Answer
Do this once, properly, in a spreadsheet. It takes under an hour and it is the difference between a decision and a guess.
- Fix your volume assumption first — One monthly minute figure, used identically for every vendor. Comparing quotes built on different volumes is meaningless.
- Convert everything to INR per month — Including FX at the rate your card actually charges, not the mid-market rate.
- Add telephony explicitly — If a vendor excludes it, get a real quote for Indian carrier costs and add it in. This line frequently changes the ranking on its own.
- Add number rental — Per number per month, multiplied by how many local CLIs you actually need.
- Add platform and seat fees — Any fixed monthly charge, regardless of usage.
- Note the tax treatment — Whether a GST invoice is issued and input credit is claimable materially changes the effective cost.
- Estimate your own integration effort — In engineering days, costed at your real rate. It does not appear on an invoice but it is money.
One question sorts most shortlists quickly: ask each vendor for a written all-inclusive monthly total at your stated volume, in rupees. The quality and directness of the answer tells you a great deal beyond the number itself.
Pricing Structures You Will Encounter in the Indian Market
The market contains genuinely different models, and each suits a different buyer. Recognising which one you are being quoted makes the comparison much faster.
| Structure | Typically offered by | Watch for |
|---|---|---|
| Per talk minute, INR, all-inclusive | India-focused platforms | Which volume tier the quoted rate assumes |
| Per minute in USD, telephony excluded | Global voice APIs | Carrier cost, FX markup, no GST credit |
| Per seat per month | Legacy contact-centre software | Cost scales with headcount, not with value |
| Platform fee plus usage | Mid-market SaaS | The fee is charged in quiet months too |
| Per call rather than per minute | Some IVR vendors | Unanswered dials may still be billed |
| Annual licence | Enterprise suites | Renews whether or not it worked |
Reconstructing the Real Cost of Any Voice AI Platform
Build this table for every vendor on your shortlist using your own projected monthly volume rather than their example numbers. It takes about twenty minutes and it routinely changes the ranking.
- Platform or licence fee — Monthly minimums, seat licences and per-agent charges. These dominate at low volume and are easy to overlook when you are focused on the per-minute number.
- Per-minute conversation rate — The headline figure — and usually the only one shown on a pricing page.
- Telephony — Whether Indian carrier costs are included or whether you must bring your own trunk. If separate, get a real quote for that too, because it is not a rounding error.
- Number rental — Per number per month. Multiplies quickly if you want local CLIs across regions.
- FX and payment costs — On USD billing, exchange rate plus a card markup, applied to every invoice.
- Tax treatment — Whether you receive a GST invoice you can claim input credit against.
- Onboarding and integration — Professional services fees, plus your own engineering time. The second is real cost even though nobody invoices you for it.
A shortcut that reveals a lot: ask each vendor for a total monthly figure at your projected volume, all inclusive, in rupees. Vendors whose pricing looks best only when telephony is excluded tend to resist answering that question directly.
Comparing the Three Billing Models Honestly
Each model suits a different situation, and pay-as-you-go is not universally best. Being clear about the trade-offs makes the choice easier to defend internally.
| Model | Best when | Weakness |
|---|---|---|
| Pay-as-you-go | Volume varies, or you are still testing | Harder to forecast an exact monthly figure |
| Committed volume tier | Volume is predictable and high | You pay for the commitment whether used or not |
| Per-seat licence | Human agents are the unit of work | Cost scales with headcount, not with value |
| Monthly platform fee | Vendor needs predictable revenue | Penalises low and variable usage |
| Per-call pricing | Calls are uniformly short | Unanswered dials still cost money |
A practical sequence: start pay-as-you-go while you are learning, then move to a committed tier once your monthly volume is stable and predictable. Committing before you know your volume is how teams end up paying for minutes they never use.
Where the Money in a Voice AI Call Actually Goes
Understanding the cost structure tells you which vendor claims are plausible and which are not. A call has four real cost components, and they are not equally significant.
- Telephony — Carrying the call over the phone network. Domestic Indian routing is substantially cheaper than routing the same call internationally, which is the largest single source of cost difference between platforms.
- Speech recognition — Converting audio to text in real time. Meaningful but not dominant, and largely fixed across vendors using comparable technology.
- Model inference — The reasoning that decides what to say. Costs vary by model choice, but most serious platforms use comparable frontier models, so this is rarely where vendors differentiate on price.
- Speech synthesis — Generating the audio. Similar across platforms.
- Everything else — Storage, dashboard, orchestration. Real but small per call, and it is where platform fees are usually justified.
The practical implication: for Indian calling, telephony routing is where the genuine cost differences live. A platform that owns Indian routes has a structural advantage that does not require cutting anything else.