AI Call Center Software Pricing: $0.05 to $0.31 a Minute

Short answer. AI call center software answers or places phone calls with a voice agent instead of a person, and most vendors that publish a price charge by the minute: Retell lists $0.07 to $0.31 a minute, Bland lists $0.12 to $0.14, and Vapi lists $0.05 a minute for hosting before the language model, voice and phone line are added. Seat-based contact center platforms bill differently, for example Twilio Flex at $150 per named user a month. For comparison, the median US customer service representative earned $21.53 an hour in 2025 according to the Bureau of Labor Statistics, which is about $0.36 a minute in wages alone. All prices below were read on the vendors' own pages on 9 November 2026.

What AI call center software is

A traditional contact center platform routes calls to human agents and records what happens. AI call center software adds a voice agent that can hold the conversation itself: it listens, decides what to say or do, and either finishes the task or hands the call to a person. Vendors sell this in two shapes. Developer platforms such as Retell, Vapi and Bland give you the voice agent and charge for usage. Contact center suites such as Dialpad and Twilio Flex give you the seats, the routing and the reporting, and add AI as a feature or a metered service.

The split matters because it decides what you pay for. A voice agent that books appointments after hours has no seats, so a per-minute price is the whole story. A team of 20 human agents with an AI assistant has seats, so a per-seat price is the base and the AI sits on top.

Three ways vendors charge

Per minute. You pay for the minutes the agent is on a call. Retell calls this "true pay as you go" and says it starts at $0. This is the easiest model to forecast if you know your call volume.

Per seat or per user. You pay for each human who logs in. Twilio Flex lists named-user pricing from $150 a month, active-user-hour pricing from $1 an hour, and a newer user-plus-usage option from $35 per monthly active user. The price tracks headcount, not call volume.

Per conversation. Dialpad says it prices its agentic AI "one conversation at a time" through a pool of AI Agent Credits, and that a conversation is billable only when the AI retrieves information or executes a real action such as scheduling, routing or an order lookup. Dialpad's pricing page shows custom pricing and no dollar figure, so we cannot give you a number.

What the vendors publish

Sources, all read on 9 November 2026: the Retell, Vapi, Bland, Dialpad and Twilio Flex pricing pages.

Read the Retell and Vapi numbers carefully. Retell's $0.07 to $0.31 range depends on which voice and model you choose: its own table lists its voice infrastructure at $0.055 a minute and ElevenLabs Flash voices at $0.040 a minute before the other parts are added. Vapi's $0.05 a minute is its hosting fee only. Its page shows example language model costs of $0.0077 to $0.0452 a minute and example voice costs of $0.0146 to $0.0238 a minute on top, and transcription and telephony are further lines we did not capture. Bland's page makes the same point about its rivals, saying that their advertised rates are the platform fee only. That is a competitor's claim, so treat it as a prompt to check each bill rather than as a finding.

A worked example at 10,000 minutes a month

This is a hypothetical, not a quote: a business whose voice agent handles 10,000 talk minutes in a month, about 167 hours. The arithmetic uses only the published rates above.

Two readings stand out. First, Bland's Build plan only beats Start once you pass 14,950 minutes a month, because the $299 fee is recovered at $0.02 a minute. Below that, Start is cheaper. Second, the human line is a floor: it counts only wages for the minutes spent talking, with no benefits, no idle time between calls, no training and no software. The Vapi and Twilio lines are also floors, because they leave out telephony and the language model or licences we could not price from the page.

Where the cost compares with a person

The Bureau of Labor Statistics puts the 2025 median pay for customer service representatives at $44,770 a year, or $21.53 an hour, across 2,666,000 jobs. Divided by 60, that is about $0.36 a minute of paid time. Retell's top rate of $0.31 a minute is 86% of that wage figure, and its bottom rate of $0.07 is about a fifth of it. The gap is real but narrower than the headline suggests once you add the parts the vendor charges separately, and a person can handle a call that is not in any script.

We would not frame this as replacing the whole team. We suggest starting with the repetitive, well-defined calls: after-hours answering, appointment booking, order status and first-pass qualification. Our comparison of an AI receptionist and a human answering service and the page on AI receptionist cost go through the same trade-off for smaller teams.

What a pricing page does not show

• Telephony. Bland's page says developer platforms bill telephony separately from the language model, speech-to-text and text-to-speech, so ask each vendor for the all-in rate.
• Language model and voice. On developer platforms these are separate lines, as the Vapi and Retell tables show.
• Concurrency. Retell includes 20 concurrent calls, Bland Start 10 and Vapi's free tier 4. Ask what happens to calls above the limit, and size the tier to your busiest hour, especially for a campaign that dials many numbers at once.
• Transfers. Bland bills transfer minutes at $0.04 to $0.05 a minute, separate from talk time.
• Build and tuning. The script, the knowledge base, the call flow and the hand-off rules are work somebody does once and keeps maintaining. None of the vendors above prices that.
• Enterprise terms. Retell, Bland, Vapi and Dialpad all show custom pricing for large deployments, so volume discounts and terms are a conversation, not a published number.

The rule that applies if the AI places the call

The software is only one part of the cost of outbound calling. In a declaratory ruling adopted on 2 February 2024 and released on 8 February 2024 (FCC 24-17), the Federal Communications Commission confirmed that the Telephone Consumer Protection Act's restrictions on an "artificial or prerecorded voice" cover current AI technologies that generate human voices. As the FCC puts it, calls that use such technologies require the prior express consent of the called party.

In practice, that means an AI agent that dials people needs a consent record before it dials, and your platform choice should include how consent, opt-outs and call recording are handled. Our guides on TCPA compliant AI calling platforms and on the TCPA checklist for an AI agent cover the details, and our comparison of AI voice agents covers the products themselves. This is general information, not legal advice.

How to choose

1. Count your minutes. Pull a month of call logs and total the talk minutes by call type. Every figure above scales from that number.
2. Pick the model that matches your shape. Spiky inbound volume with few humans suits per-minute pricing. A staffed team with an AI assistant suits per-seat pricing.
3. Ask for the all-in rate. Get the per-minute price including model, voice, transcription and telephony in writing, not just the platform fee.
4. Test concurrency. Check the call limit against your busiest hour.
5. Pilot on one call type. Run it for a month with a human fallback and compare the cost per completed call with your current figure.

How imisofts can help

We build and configure AI voice agents on the platforms above and on others, connect them to a CRM or calendar, and set up the hand-off and consent steps. See our page on AI calling built around TCPA consent, or the AI calling for home services page if your calls are about bookings and quotes. For a related buying guide, read AI answering service for HVAC.