How much does an AI voice agent cost for a small business? Pricing broken down.

What an AI voice agent actually costs per minute

There's no single "AI voice agent price." Every platform bills by the minute of talk time, and that per-minute rate is actually a stack of separate charges bundled into one number: a platform or orchestration fee, a speech-to-text and text-to-speech fee, an LLM (the language model that decides what to say) fee, and usually a telephony fee for the phone call itself.

Across the three platforms with public pricing (Retell, Vapi, and Bland), all-in rates run from roughly $0.07 to $0.31 per minute depending on the platform and voice model, before you add the phone call. Retell's own worked example lands at $0.11 per minute for a typical setup: $0.04/min for the LLM, $0.055/min for voice infrastructure, and $0.015/min for text-to-speech. That's the number to anchor on if you want a quick gut-check, but the plan fees, concurrency limits, and telephony add-ons underneath it vary enough to change your actual monthly bill by hundreds of dollars. The rest of this guide breaks those apart.

Retell, Vapi, and Bland pricing compared

Here's what each platform actually charges, as published on their pricing pages.

Retell and Vapi both separate the AI costs into components (voice infrastructure, speech-to-text/text-to-speech, LLM), while Bland quotes one blended per-minute talk rate plus a separate transfer-time rate for calls handed off to a human. Bland's Build plan also raises your concurrency ceiling to 50 simultaneous calls and 2,000 calls a day, versus 10 concurrent calls and 100 calls a day on Start, a relevant difference if you're a multi-location business rather than a single front desk.

Concurrency and extra phone numbers are where these platforms quietly add cost. Retell charges $8/month per concurrent line beyond its 20 free slots, plus $2/month for a Retell-hosted number. Vapi charges $10/line/month beyond plan limits and $20/org/month for additional organizations. If you're only ever handling one call at a time, none of this matters; if you're a growing office fielding simultaneous calls during peak hours, it adds up fast. Our guide on after-hours answering options for contractors walks through a related concurrency problem: calls that pile up outside business hours.

The hidden line item: telephony

None of the three platforms' per-minute AI rates include the actual phone call unless you use their bundled number, and even then it's billed separately under the hood. The phone call itself routes through a carrier like Twilio, and Twilio bills it on its own schedule: $0.014/min for outbound calls to US local numbers, $0.0085/min for inbound calls to US local numbers, and $0.022/min for inbound calls to toll-free numbers. A Twilio local number runs $1.15/month; a toll-free number runs $2.15/month.

Retell's own breakdown shows this explicitly: telephony costs $0.015/min when you use a Retell number through Twilio or Telnyx, but it's free if you bring your own SIP trunk (a direct connection to a telephony provider that bypasses the per-minute carrier fee). Vapi's own transport (SIP, WebSocket, or Daily WebRTC) is free, but routing calls through Twilio still costs $0.008/min inbound and $0.014/min outbound on top of whatever Vapi charges for the AI stack.

The practical takeaway: when you see a vendor quote a single per-minute number, ask whether it includes the phone call or just the AI processing. If it's just the AI processing, add telephony on top. If you want call recording or transcription as a backup record, Twilio adds $0.0025/min for recording, $0.0005/min for storage, and $0.05/min for transcription.

What 1,000 minutes a month really costs

The vendors themselves publish worked examples for 1,000 minutes of monthly call volume, which is a useful stand-in for a business getting a steady but not huge amount of phone traffic.

Retell's own math: at its $0.11/min blended rate, 1,000 minutes costs about $110 for the AI stack alone, before telephony. Add Retell's own telephony rate of $0.015/min for a Retell number routed through Twilio or Telnyx, and the per-minute total climbs from there; that telephony charge drops to zero if you bring your own SIP trunk instead.

Vapi's own worked example for the same 1,000 minutes comes out to $82 to $129 a month total: $50 hosting, $10 for the transcriber, $8–$45 for the intelligence model, and $15–$24 for the voice model. That range already includes the components Retell lists separately, which makes a side-by-side comparison a little messy. Both platforms land in a broadly similar range for moderate call volume, before carrier telephony charges are added.

Bland's Start plan, by contrast, is pure usage: 1,000 minutes of pure talk time at $0.14/min comes to $140, with no platform fee attached. That's the highest per-minute rate of the three, but it has no fixed monthly cost sitting underneath it, which matters if your volume is unpredictable month to month.

None of these totals are guarantees for your specific call mix. Real calls include hold time, transfers, and silence, and all of that gets billed. They're still the only verified reference points the vendors publish, close enough to budget against.

AI voice agent vs. a human receptionist

The other side of the cost question is what you're comparing the AI agent against. Federal labor data gives a solid benchmark here. The median hourly wage for receptionists was $18.27/hr as of BLS's May 2025 data, with a median annual wage of $38,010.

But wages are only part of what an employer actually pays. BLS's June 2026 compensation data shows wages and salaries make up 68.4% of total civilian worker compensation, with benefits making up the other 31.6%. Applying that same benefits share to the receptionist wage gives an estimated fully loaded cost of roughly $26.71/hr, a calculation combining those two BLS figures rather than a single published number, but a reasonable way to compare apples to apples.

That $26.71/hr figure buys continuous coverage for one line, one caller at a time, during the hours someone is actually at the desk. An AI voice agent's cost scales with minutes actually talked, not hours staffed, and it can answer multiple calls at once without the per-line staffing cost scaling the same way. BLS also projects receptionist employment to decline about -2% from 2025 to 2035, even as roughly 105,100 openings are projected per year from people leaving the field, a sign the hiring pool for this role is already tight in a lot of markets. That matters if your real alternative to automation is hiring rather than sticking with what you have now.

None of this means AI is automatically less expensive. A receptionist also handles front-desk tasks a phone bot can't, like greeting walk-ins and managing email; a fair comparison has to account for what the role actually covers beyond answering calls. For a deeper look at that specific trade-off, see our guide on AI answering service vs. virtual receptionist.

When the fixed fees aren't worth it

Fixed monthly platform fees can flip the math against automation at low call volume. Vapi's Core plan costs $29/month regardless of how many minutes you actually use, and its Pro plan has a $999/month minimum regardless of usage; a business fielding only a handful of calls a month would pay far more in platform fees than in actual usage charges on that tier.

Bland's Build plan adds a $299/month fee on top of per-minute talk time, and that fee only pays for itself once your volume needs its higher caps: 50 concurrent calls and 2,000 calls a day, versus 10 concurrent calls and 100 calls a day on the free Start tier.

Healthcare, dental, and legal intake use cases carry an extra cost most owners don't see coming: HIPAA-compliant configuration. Vapi charges $2,000/month for HIPAA compliance on top of its regular usage fees, which is a meaningful line item for a small practice weighing this against a receptionist or an answering service. Our guide on AI automation for medical practices covers that trade-off in more depth for clinical settings.

The general rule: if your call volume is low and steady, a pay-as-you-go plan with no monthly minimum (Retell's pay-as-you-go tier or Bland's Start plan) avoids paying for capacity you don't use. If your volume is high and you need concurrency or compliance features, the fixed-fee plans start to make more sense because the per-minute rate usually drops as the platform fee rises.

What drives the final number for your business

Four variables determine what you'll actually pay, and none of them show up in a vendor's headline rate:

• Call volume. Pay-as-you-go rates (Retell's $0.07–$0.31/min, Bland's $0.14/min Start rate) favor low, unpredictable volume; fixed platform fees (Vapi's $29–$999/month, Bland's $299/month) favor steady, higher volume where the per-minute rate effectively drops.
• Concurrency. If calls ever stack up at the same time, check the free concurrent-call limit (20 on Retell, 10 on Vapi Core, 10 on Bland Start) against your real peak, not your average.
• Telephony. Decide whether you're using the vendor's bundled number (simpler, billed at their telephony rate) or a separate carrier like Twilio (potentially lower per minute, more moving parts).
• Compliance. HIPAA or other regulated-industry requirements add a flat fee on top of usage, as Vapi's $2,000/month HIPAA charge shows.

Run your own numbers against your actual call logs before picking a platform, not against the headline per-minute rate. For the broader question of whether an AI phone system is the right fit for your business at all, our guide on AI receptionists for small business walks through the decision from the start, and the AI Voice Agent service page covers how an AI voice agent answers calls, books jobs, and sends text summaries.