Research firms put the AI voice agents market size at about USD 2.4 to 3.4 billion in 2024 and 2025, and forecast USD 35 to 55 billion by 2033 to 2036, growth of roughly 24% to 39% a year. These are forecasts, and they differ because each firm defines the market differently.


“We're ShiftMate, a South African voice agent specialist, and we read the market forecasts as a company building in this market. This guide covers the AI voice agents market size, how the forecasts differ, who's investing, and the shift we think the headline numbers miss: who is buying, and why.”
The AI voice agents market size was about USD 2.4 to 3.4 billion in 2024 and 2025, according to four research firms, and they forecast it to reach USD 35 to 55 billion between 2033 and 2036. AI voice agents are software that hold phone and voice conversations: answering, booking, selling and supporting. If that's new to you, our guide to AI voice agents explains them first.
Here are the four published AI voice agents market size estimates side by side, with the period each growth rate covers. No firm gives its full method without the paid report, so treat them as informed forecasts.
| Research firm (published) | Recent size | Forecast | Growth a year (CAGR) |
|---|---|---|---|
| Grand View Research (June 2026) | USD 2.5 billion (2025); USD 3.5 billion estimated for 2026 | USD 35.2 billion by 2033 | 39.0% (2026 to 2033) |
| Market.us (August 2025) | USD 2.4 billion (2024) | USD 47.5 billion by 2034 | 34.8% |
| Meticulous Research, sold by Research and Markets (October 2026) | USD 3.40 billion (2025); USD 5.10 billion estimated for 2026 | USD 43.83 billion by 2036 | 24.0% (2026 to 2036) |
| SNS Insider (2026) | USD 2.75 billion (2025) | USD 55.55 billion by 2035 | 35.1% (2026 to 2035) |
Grand View Research adds that North America held the largest share, 38.1% of the market, in 2025. We found no public figure for Africa or South Africa in any of the four releases.
AI voice agents market size estimates differ because the firms count different things, start from different years and measure growth over different periods. Three things to check before you quote any of them:
The October 2026 report counts a lot: agent platforms, speech recognition, synthetic voices and voice cloning, speech-to-speech models, telephony and contact centre connections, analytics, testing and QA, set-up and managed services, across phone, web, apps, drive-throughs, kiosks and cars (release on Yahoo Finance). A narrower definition gives a smaller number.
Market.us starts from 2024; the others from 2025. A fast-growing market looks very different one year apart.
A 24% rate over ten years (2026 to 2036) and a 39% rate over seven (2026 to 2033) can land in a similar place. Compare end values and end years, not growth rates alone.
Watch for copies. A November 2026 release forecasting USD 35.24 billion by 2033 matches the Grand View figure, so it isn't an independent estimate. And we couldn't trace a widely repeated claim that the voice AI market reached USD 22.5 billion in 2026 to any research firm, so we don't use it.
| Market | Firm (published) | Figure | Status |
|---|---|---|---|
| Conversational AI (chat and voice) | MarketsandMarkets (April 2024) | USD 13.2 billion (2024) to USD 49.9 billion by 2030, 24.9% a year | Forecast |
| Conversational AI (chat and voice) | Grand View Research (June 2022) | USD 41.39 billion by 2030, 23.6% a year | Older forecast |
| Speech and voice recognition | MarketsandMarkets (August 2025) | USD 9.66 billion (2025) to USD 23.11 billion by 2030, 19.1% a year | Forecast |
| Voice recognition | Grand View Research (2026) | USD 27.2 billion (2025) to USD 80.8 billion by 2033, 14.6% a year | Forecast |
| Conversational AI in contact centres (spending) | Gartner (August 2022) | USD 1.99 billion end-user spending in 2022 | Estimate |
Every conversational AI market growth forecast here is larger than the voice agents figures because it includes chatbots, or every use of speech recognition. The two speech recognition forecasts disagree by almost three times for the same year, which shows how much the definition matters.
Analytics, testing and quality checking is the fastest-growing part of the voice AI market, according to the October 2026 Research and Markets release, because buyers now put compliance, performance monitoring and safety first. Agent platforms remain the largest part (release). Every AI voice agents market size forecast above is a total; the segments show where the money goes inside it. The release's segment picture:
| Segment | Largest | Fastest-growing |
|---|---|---|
| Component | Agent platforms | Analytics, testing and QA |
| How it's built | Cascaded (speech to text, then a language model, then speech) | Speech-to-speech |
| Call direction | Inbound | Inbound (outbound stays limited to consented reminders, follow-ups and collections) |
| Channel | Phone and contact centre | Web and in-app voice |
| Use | Customer service | Patient engagement |
| Buyer size | Large enterprises | Small and medium businesses |
| Industry | Banking, financial services and insurance | Healthcare |
| Region | North America | Asia-Pacific |
| Deployment | Cloud | Not stated |
The testing finding matches what we see. Building a voice agent that sounds good in a demo is no longer the hard part; proving it behaves on every kind of call is. That's why our AI sales agents have to pass their practice situations before they phone anyone. Our page on why AI voice agents fail explains what that testing involves.
“Before a version of one of our AI sales agents may phone anyone, it has to pass every practice situation three times: the busy prospect, the one who only wants an email, the one who asks if it's a robot, the one who says take me off your list. Passing once proves nothing.”
Faster, more natural speech models and the demand for round-the-clock service are driving the market; regulation, fraud, running costs and customer trust are holding it back. Both lists come mainly from the October 2026 release (release), with our notes from building agents. These forces explain why every AI voice agents market size forecast assumes fast growth, and why the range between them is so wide.
| Driving growth | Holding it back |
|---|---|
| Real-time speech models with shorter delays, better handling of interruptions, more expressive voices and more languages | Rules on AI calls: the US FCC ruled on 8 February 2024 that AI voices count as "artificial" under the TCPA, so callers must have given prior express consent, and written consent for telemarketing |
| Contact centres wanting service at any hour, and lower cost per call | Fraud: the release cites US FTC figures of USD 12.5 billion in consumer fraud losses in 2024, about USD 2.95 billion from impostor scams |
| Large existing spend on call centres and outsourcing to replace | Transparency duties under the EU AI Act, which the release says apply from August 2026 |
| Agents that complete tasks (bookings, payments, claims), not just answer questions | Running costs: Gartner predicts AI cost per resolution will pass USD 3 by 2030 |
| Multilingual agents for emerging markets (named as an opportunity) | Accuracy across accents and dialects (named as a challenge), and many customers still preferring a person |
The release also names what suppliers compete on: speed of reply, voice quality, accuracy across languages, connections to the systems a business already uses, compliance, reliability and price per minute.
“A rule we wrote for outbound calls broke inbound ones: it flagged a perfectly normal caller saying "Hi, I'm phoning about my fibre" as a mistake. The same sentence is right on one kind of call and wrong on another, so we test every call type on its own.”
Venture investors and large software companies are both putting money in, betting that the AI voice agents market size forecasts are right. CB Insights counted voice AI funding of USD 394 million in 2021, USD 315 million in 2022 and USD 264 million in 2023, then USD 2.1 billion in 2024 and nearly USD 500 million in the first quarter of 2025 (CB Insights, May 2025). Andreessen Horowitz argued in January 2025 that the most natural early categories for voice agents are those with high existing call centre and outsourcing spend (a16z).
| Company (what it does) | Deal | Date | Source |
|---|---|---|---|
| Sierra (customer service agents, chat and voice) | USD 950 million raised, valued at over USD 15 billion | May 2026 | TechCrunch |
| Wonderful (enterprise AI agents, began with voice support) | USD 550 million Series C, USD 5 billion valuation | September 2026 | Release |
| Parloa (AI customer service agents, Berlin) | USD 350 million Series D, USD 3 billion valuation | January 2026 | Reuters |
| Decagon (customer service agents) | USD 250 million Series D, USD 4.5 billion valuation | January 2026 | Business Wire |
| PolyAI (enterprise voice agents, London) | USD 86 million Series D, about USD 750 million valuation | December 2025 | Forbes |
| NICE buys Cognigy (conversational and voice AI) | USD 955 million acquisition, closed | September 2025 | Cognigy |
| SoundHound AI to buy LivePerson | Implied enterprise value of USD 250 million | April 2026 | SoundHound |
| Salesforce to buy Cresta (voice-first customer experience) | Price not disclosed | November 2026 | Salesforce |
The pattern: the biggest money is going to agents that handle customer service end to end, and established customer service software companies are buying voice AI companies. The same a16z update notes buyers paying for voice agents that act as simulators for coaching and training, a use outside the classic call centre.
Analysts predict AI will resolve more routine customer service, but not replace people, and not always cost less. These are Gartner's predictions and survey findings, in date order:
Gartner predicted that conversational AI in contact centres would reduce agent labour costs by USD 80 billion by 2026, and that one in ten agent interactions would be automated by 2026, against about 17 million contact centre agents worldwide. We found no Gartner statement on whether it came true. Source.
Of 187 customer service leaders, 44% were exploring a customer-facing AI voicebot, 11% piloting one and 5% had one deployed. Source.
Agentic AI will autonomously resolve 80% of common customer service issues without human intervention by 2029, cutting operational costs by 30%. Source.
By 2028, none of the Fortune 500 will have fully eliminated human customer service, and by 2027 half of organisations that planned big service-staff cuts because of AI will drop those plans. Source.
By 2030, cost per resolution for generative AI will exceed USD 3, higher than many offshore human agents, and AI regulation will raise assisted service volume by 30% by 2028. Source.
Read together with the AI voice agents market size forecasts, they point to a hybrid: AI takes the routine calls, people take the rest, and the remaining human calls get harder.
For a single business, the market totals matter less than the unit price. Published per-minute prices depend on what they include: one developer platform lists US$0.07 to US$0.31 a minute, built from itemised parts (voice engine, voice, AI model and a US$0.015 phone line), while hosting-only building blocks start at about US$0.01 a minute and bill the AI models and phone line separately (vendors' own pricing pages, checked 16 November 2026; see AI voice agent pricing).
We found no public AI voice agents market size figure for South Africa in any of the four releases. The October 2026 report lists South Africa as a covered country, but its release gives no figure, so we won't invent one. What we can measure is the industry voice AI will change: South Africa's global business services sector, call centres and back offices serving international clients, grew from 65,000 people in 2019 to an estimated 150,000 in 2024, with the UK accounting for 55% of that headcount and the US 33% (BPESA and Everest Group estimates, March 2025).
AI is already changing the work. A South African BPO told The National its AI collections agent dealt with more than 100,000 people a day, and a large operator said AI had stretched agent training from three weeks to eight (The National, November 2026). The report's named opportunity, multilingual agents for emerging markets, and its named challenge, accuracy across accents and dialects, are exactly South Africa's situation. Our guide to AI voice agents in South Africa covers accents, POPIA, the opt-out registry and rand pricing.
“The market responds better to natural, authentic South African voices. An American or British voice on a South African line sounds like a scam call. So our sales agents always sound South African, from the first word to the last, even on UK or US campaigns.”
Voice AI is heading towards agents that complete tasks, sound local, are tested before they speak, and work alongside people. Here is what we expect, and what no market forecast can tell you about your own project.
A growing AI voice agents market size doesn't make a given agent reliable. McKinsey writes that most voice agent implementations stay at the pilot stage, mainly because they aren't reliable enough for real customers (McKinsey, June 2026).
The reports are paid products. The public releases give chapter names, not methods, so nobody outside can check the numbers.
Market totals say nothing about your cost per resolved call in rand. Price your own calls.
A forecast can't hear whether an agent understands your callers' accents. Only testing on your own calls can.
The forecasts above count spend on platforms and minutes. We think they miss two things. The first is who is buying: a practice or a small sales team that never had a call centre isn't replacing agents, it's replacing voicemail and missed calls, and that shows up in no headcount figure or forecast of agent jobs. The second is what buyers judge it on: the lead qualified or the meeting booked, not minutes or seats. A forecast in minutes can't tell you whether those results happen.
That's the market we build for: our AI outbound sales agent exists to qualify leads and book meetings, and our AI practice calls to get people ready for the floor.
If you're sizing this market for your own business, start with your own numbers: the calls you miss and the calls nobody makes. Send us those and we'll size your share of it in rand.
“Voice AI isn't replacing call centres. What it's doing is making voice available and affordable to businesses that could never afford a call centre, and supporting the call centres that exist: training their people, generating leads, and making the low-value calls their operators don't want to make. This space is changing and opening up.”
“Everything we build runs on one engine: the same voice AI that sells also trains people and interviews candidates, so every lesson from one product makes the others better. That engine, built in South Africa and tested on South African calls, is how we intend to lead voice AI here, and what we intend to take to international markets.”
Research firms put the AI voice agents market size at about USD 2.4 to 3.4 billion in 2024 and 2025, and forecast USD 35 to 55 billion by 2033 to 2036, growth of about 24% to 39% a year. Grand View Research, Market.us, Meticulous Research and SNS Insider each define the market differently, so their figures differ.
MarketsandMarkets forecast the conversational AI market, covering chat and voice, to grow from USD 13.2 billion in 2024 to USD 49.9 billion by 2030, about 24.9% a year. It's larger than the voice agents market because it includes chatbots.
Analytics, testing and quality checking, according to Research and Markets in October 2026, as buyers focus on compliance, performance and safety. Agent platforms remain the largest part. Inbound calls are the largest and fastest-growing use, and healthcare the fastest-growing industry.
Published per-minute prices depend on what they include: one developer platform lists US$0.07 to US$0.31 a minute, built from itemised parts (voice engine, voice, AI model and a US$0.015 phone line), and hosting-only building blocks start at about US$0.01 a minute (checked 16 November 2026). Gartner predicts AI cost per resolution will exceed USD 3 by 2030.
Quoting one firm's figure as fact, comparing growth rates over different years, mixing voice agent figures with wider conversational AI or speech recognition markets, repeating figures with no traceable source, and assuming a growing market means a given project will work. Name the firm, the year and the definition every time.
Analysts expect AI to handle a growing share of routine calls, but not to remove people. Gartner predicts that by 2028 no Fortune 500 company will have fully eliminated human customer service, and that by 2030 AI cost per resolution will exceed USD 3.
Not with a South African figure. The October 2026 Research and Markets report lists South Africa as a covered country, but we found no public figure for South Africa in its release. The best local data is BPESA and Everest Group's estimate of about 150,000 people in global business services in 2024.
CB Insights counted USD 2.1 billion of voice AI venture funding in 2024, up from USD 264 million in 2023. The largest 2026 customer service agent rounds ran from USD 250 million to USD 950 million, and established customer service software companies bought voice AI companies.
Start with the calls you miss and the calls nobody makes, not with a market forecast. That's where we think voice AI pays first.