How Much Does an AI Voice Agent Cost?
Content

Most AI voice agents are billed per minute of conversation. In November 2026, the published usage rates of well-known voice AI platforms start at roughly $0.07 per minute and can go above $0.30 per minute, depending on the language model, the voice, the telephony and the add-ons you choose. On top of that you usually pay fixed monthly items (a platform fee or plan, phone numbers, extra concurrent calls) and a one-off setup effort to connect the agent to your calendar, CRM and processes. For a small business handling a few thousand minutes a month, the realistic question is not “what is the price per minute?” but “what will the whole workflow cost, and what will it save?”
This guide breaks down every cost component, compares the three main pricing models, shows published prices from several platforms, and gives you a simple formula to estimate your own bill. Prices were checked on vendors’ public pages on 16 November 2026; they change often, so always confirm them before you sign.
If you are new to the topic, start with what an AI voice agent is and how an AI voice agent works.
What you actually pay for
An AI voice agent is not a single product line on an invoice. Behind every call there is speech recognition, a language model, a synthetic voice, a phone line and the platform that orchestrates them. Some vendors bundle all of this into one rate; others bill each piece separately.
| Cost component | What it covers | How it is usually billed |
|---|---|---|
| AI usage | Speech-to-text, language model (LLM), text-to-speech and orchestration | Per minute of call |
| Telephony | Carrier cost of the inbound or outbound call | Per minute, by country |
| Phone numbers | Local, mobile or toll-free numbers for the agent | Per number, per month |
| Platform or plan | Access to the software, support level, data retention | Monthly fee, or none on pure pay-as-you-go |
| Concurrency | How many calls the agent can handle at the same time | Included up to a limit, then per extra line |
| Add-ons | Knowledge base, noise reduction, PII redaction, quality assurance, branded caller ID | Per minute, per call or per month |
| Follow-up messages | SMS, WhatsApp or email sent after the call | Per message |
| Setup and integration | Designing the conversation, connecting calendar and CRM, testing | One-off, internal time or partner fee |
| Supervision | Reviewing transcripts, improving prompts, handling escalations | Your team’s time |
The first three lines are what most “price per minute” comparisons show. The last three are where budgets usually go wrong.
The three common pricing models
1. All-inclusive per-minute rate
One rate per minute covers the language model, speech recognition and voice. It is the easiest model to forecast: minutes × rate, plus telephony and numbers if they are billed separately. The trade-off is less control over which model or voice runs underneath.
2. Platform fee plus pass-through components
You pay the vendor a per-minute platform or hosting fee, then pay separately (directly or passed through at cost) for the transcriber, the LLM, the voice and the phone line. This gives developers maximum flexibility to pick models, but the final rate per minute depends on every choice you make, and the stack has to be built and maintained by someone technical.
3. Subscription plus usage inside a conversation platform
The voice agent is part of a broader customer platform or CRM. You pay a monthly plan and then usage (minutes, messages) through credits or a per-minute rate. Here the comparison is less about the voice minute and more about what else is included: the customer record, the other channels, automations and integrations you would otherwise have to buy and connect.
Large companies often end up on a fourth option, a custom enterprise contract with volume pricing, dedicated infrastructure and SLAs.
What published prices look like in 2026
The table below summarises the public pricing pages of three developer-oriented voice AI platforms, as published on 16 November 2026, in US dollars and excluding taxes. It is not a ranking and it does not measure quality, latency or features; it shows how differently the same “per minute” can be built.
| Platform | Pricing model | Published usage price | Fixed or extra items (examples) |
|---|---|---|---|
| Retell AI | Pay-as-you-go, components | $0.07–$0.31 per minute in total, depending on LLM, voice and telephony; voice infrastructure alone is $0.055/min | Phone numbers $2/month; 20 concurrent calls included, then $8 per extra concurrent call per month |
| Vapi | Hosting fee plus providers at cost | $0.05/min hosting fee, plus transcriber, model and voice billed at provider cost | Optional packages: Core $29/month; Pro at 10% of the hosting fee with a $999/month minimum. Vapi’s own calculator estimates $82–$129 for 1,000 minutes a month |
| Bland AI | All-inclusive per minute (LLM, STT and TTS) | Start: $0.14/min with no platform fee; Build: $0.12/min | Build plan has a $299/month platform fee; transfer minutes $0.05 or $0.04/min; telephony billed separately |
Three things stand out. First, the cheapest-looking headline rate is rarely the full rate: on component-based platforms you add the model, the voice and the phone line. Second, fixed fees matter at low volume and fade at high volume. Third, these platforms are toolkits for building an agent; the CRM, the calendar, the follow-up on other channels and the reporting are usually separate projects. For a feature comparison rather than a price comparison, see our practical comparison of AI voice agents for business in 2026 and Spoki vs Retell vs Vapi.
Hidden costs that change the bill
- Silence and hold time. Many providers bill the whole connected call, including pauses, because speech recognition keeps listening. Long holds and slow back-end lookups cost money.
- Minimum billed time and long prompts. Some vendors apply a minimum per call or bill extra time when the agent’s prompt and knowledge are very large. Read the billing rules, not just the rate.
- Transfers. When the AI hands over to a person, the AI fee may stop, but telephony usually continues for the rest of the call.
- Premium models and voices. Switching to a larger LLM or a premium voice can multiply the per-minute cost. Test whether callers actually notice the difference.
- Peaks and concurrency. If ten customers call at once and your plan allows four simultaneous calls, the rest wait or fail. Extra concurrency is often a monthly fee.
- Integration and maintenance. Connecting the agent to your calendar, CRM, ecommerce or help desk, and keeping those connections working, is often the largest cost in year one if it is custom-built.
- Compliance. In the EU you must tell callers they are speaking to an AI (Article 50 of the AI Act applies from 2 August 2026), handle personal data under the GDPR and respect consent for outbound calls. Data residency, DPAs and call recording settings can affect which plan you need.
- What happens after the call. A confirmation by SMS or WhatsApp, a summary in the CRM, a ticket for the team: each step either costs money on another tool or is included in the platform.
How to estimate your monthly cost
You can get a useful first estimate with four numbers:
- Calls per month the agent will handle (inbound, outbound or both).
- Average call length in minutes. Booking and FAQ calls are often short; qualification and support calls run longer.
- All-in cost per minute: AI usage + telephony + add-ons.
- Fixed monthly items: plan or platform fee, phone numbers, extra concurrency.
Monthly cost ≈ (calls × average minutes × all-in rate per minute) + fixed monthly items. Then add the one-off setup separately.
Here is an illustrative example. These are not quotes from any vendor; they only show how quickly volume and rate interact. Assume 2,000 calls a month at 3 minutes each, which is 6,000 minutes.
| All-in rate per minute (illustrative) | Usage per month | With €100/month of fixed items |
|---|---|---|
| €0.10 | €600 | €700 |
| €0.15 | €900 | €1,000 |
| €0.25 | €1,500 | €1,600 |
The difference between a “cheap” and an “expensive” minute is real, but so is the difference between a 3-minute and a 5-minute call. Shorter, well-designed conversations, fast answers and fewer unnecessary transfers often save more than switching vendors.
Cost versus value: what to compare it with
A per-minute price only makes sense next to what the agent changes. Before you buy, measure two weeks of baseline:
- Missed calls and how many are never called back.
- Calls outside opening hours that today go to voicemail.
- Minutes your team spends on repetitive calls (opening hours, order status, bookings, rescheduling).
- Leads that wait before someone qualifies them.
- Bookings and no-shows, if appointments are part of the business.
Then compare the agent’s monthly cost with the conversations recovered and the hours freed. In a veterinary clinic case published on this blog, missed calls fell from 68% to 4% after an AI voice agent started answering; the same article reports 35% fewer no-shows and 40% less front-desk time (AI voice agents for clinics). Those figures are Spoki-reported, not an independent study, but they show which numbers to track. For the staffing side of the comparison, read AI receptionist vs human receptionist. The right model is rarely “AI instead of people”; it is AI on the repetitive calls, with a clean handover to a human when it adds value.
Questions to ask any vendor before you sign
- What exactly is included in the per-minute rate: LLM, speech recognition, voice, telephony?
- Is billing per second, and is there a minimum per call?
- Are silence, hold time and transfers billed, and at what rate?
- How many simultaneous calls are included, and what does one more cost?
- What do phone numbers cost in each country you need? Can you bring your own numbers or phone system?
- Which integrations are native (calendar, CRM, ecommerce, help desk), and which need custom development?
- Can the agent continue the conversation by SMS, WhatsApp or email, and how are those messages billed?
- Where is data processed and stored, is a DPA available, and how is the AI disclosure handled?
- What does support cost, and who maintains the agent after launch?
- Can you test it on your own use case before committing?
For a broader buying checklist, see how to choose the right AI voice agent for your business.
How Spoki Voice is priced
Spoki is the agentic conversational CRM. It brings voice, WhatsApp, SMS and email together in a single thread per customer. Its AI agents answer calls and messages instantly, qualify opportunities, book appointments and keep the CRM up to date, so your team steps in only when it adds value. Spoki helps companies around the world connect with customers at every stage of the journey, from marketing to sales and support.
Spoki Voice is the AI voice agent inside that CRM. It answers inbound calls and makes outbound calls with end-to-end latency under 600 ms, and noise cancellation lets it hold calls even in very noisy environments. During a call it can segment contacts, qualify leads, book appointments, create orders and provide customer support. Voice can be bought standalone, without WhatsApp, and is still connected to the whole Spoki CRM and every channel, so the conversation can continue on WhatsApp, SMS or email in the same history. Spoki connects to 4,000+ tools and is an official Meta Business Partner.
On pricing, Spoki follows the third model above. Its public pricing page combines a monthly plan (from €39/month with annual billing, excluding VAT, as listed on 16 November 2026) with prepaid credits that pay for channel usage, AI and phone numbers. Voice minutes are billed per minute, as a per-country call rate plus an AI fee when the agent handles the call; the first phone number is included and additional lines are €4.99/month. Check the pricing page for the current per-country rates, because they vary by destination.
When you compare this with a developer platform, compare the whole workflow: the voice minute, plus the CRM record, the follow-up channels, the integrations and the time your team would spend building and maintaining them.
FAQ
How much does an AI voice agent cost per minute?
Published usage rates in November 2026 typically range from about $0.07 to more than $0.30 per minute, depending on the language model, voice, telephony and add-ons. Check whether the quoted rate is all-inclusive or a platform fee to which other components are added.
Are there monthly fees on top of the per-minute price?
Usually yes: a plan or platform fee (on some vendors), phone numbers, extra concurrent calls and optional add-ons. Pure pay-as-you-go plans can start at $0 per month, but numbers and telephony are still billed.
How much does it cost to set up an AI voice agent?
It depends on how many workflows you automate and how many systems the agent has to connect to. A simple FAQ or booking agent on a no-code platform can be configured quickly by your own team; custom integrations with CRM, ecommerce or legacy phone systems take longer and may need a partner.
Is an AI voice agent cheaper than a human receptionist?
It covers different things. An agent can answer every call 24/7 and handle peaks, while people handle complex, emotional or unusual cases. Compare the agent’s cost with missed calls, after-hours demand and hours spent on repetitive calls, not only with a salary.
Why do two quotes with the same per-minute price end up different?
Because of what is excluded: telephony, numbers, concurrency, silence and transfer billing, minimums per call, premium voices, and the follow-up messages and integrations around the call.
How can I reduce the cost of an AI voice agent?
Design short, focused conversations, give the agent fast access to the data it needs, avoid unnecessary transfers, choose a model and voice that are good enough for the task, and use the channels that are cheapest for each step, for example a WhatsApp or SMS confirmation instead of a second call.
Can I buy Spoki Voice without WhatsApp?
Yes. Spoki Voice can be bought standalone and stays connected to the Spoki CRM and to other channels if you add them later.
Conclusion
An AI voice agent costs a per-minute usage rate plus a handful of fixed and one-off items, and the total depends far more on your call volume, call length and integrations than on the headline rate. Estimate your minutes, ask every vendor what the rate includes, measure a baseline, and judge the project on conversations recovered and hours saved.
Want to see what it would cost and do on your own calls? Request a demo and we will walk through your use case, or try Spoki Voice for free first.
Sources (checked 16 November 2026)
- Retell AI pricing
- Vapi pricing
- Bland AI pricing
- European Commission, transparency obligations under Article 50 of the AI Act
- Spoki pricing and Spoki Voice pages
- Spoki Voice latency (under 600 ms) and noise cancellation: confirmed by Spoki
- Spoki-reported clinic figures





