AI phone agents: which startup is ahead?

Which startups are really competing in AI phone agents?

Eight startups currently deserve a place in a direct AI phone-agent comparison: Vapi, Retell AI, Synthflow, Parloa, Bland, PolyAI, Replicant and Avoca.

We focus on companies whose core product can answer or place phone calls, understand what the caller wants, connect to business systems and complete tasks without a human handling every step. The group covers developer platforms, no-code products, managed enterprise systems and one major vertical specialist.

We exclude speech suppliers such as Deepgram and Cartesia because they provide pieces of the technology rather than complete phone agents. ElevenLabs is the most important borderline case, but its public numbers still combine voice generation, media tools and conversational agents. Sierra and Decagon also support voice, although they primarily sell broader customer-service agents across several channels. Comparing their company-wide revenue with phone-first startups would create a false ranking.

The latest disclosed funding totals range from roughly $5.1 million for Retell to more than $560 million for Parloa. These figures come from company announcements, TechCrunch reporting and current company profiles. They show how differently the competitors have been built, but funding alone tells us very little about who is winning.

Startup	What it does	Cumulative funding
Vapi	Developer platform for building highly customized voice agents	About $72M
Retell AI	Usage-based platform for inbound and outbound phone agents	About $5.1M
Synthflow	No-code and managed voice-agent platform	About $30M
Parloa	Managed AI agents for large customer-service operations	More than $560M
Bland	Full-stack voice platform with proprietary infrastructure	More than $100M
PolyAI	Enterprise conversational agents for large contact centres	More than $200M
Replicant	Contact-centre automation for complex service calls	About $112M
Avoca	AI front office for home and field-service businesses	More than $125M

Is there a clear AI phone-agent leader today?

Vapi is ahead today, with Retell close behind and Synthflow emerging as the strongest third contender.

Vapi has the broadest set of advantages. It reports more than one billion calls handled, between one million and five million calls per day, more than one million developers and a recurring-revenue run rate somewhere in the “healthy eight figures.” TechCrunch also reported that Amazon Ring compared more than 40 voice vendors before choosing Vapi and now sends all its inbound calls through the platform. That combines large usage, strong distribution and a demanding enterprise customer in a way no rival currently matches.

Retell is close because its commercial numbers are unusually strong. The company says it reached $50 million in annual recurring revenue while processing more than 50 million calls each month, despite raising only about $5.1 million. Synthflow currently claims more than 65 million monthly calls across over 30 countries, giving it comparable usage to Retell, although it has disclosed much less about revenue and major customer deployments.

Parloa leads a different race. It sells to large international companies, passed $50 million in recurring revenue and raised enough money to expand aggressively in the United States and Europe. PolyAI has the deepest disclosed enterprise deployment base, while Bland controls more of its own voice infrastructure. Neither currently matches the top three across enough dimensions to take the overall lead.

Our current view puts Vapi at 100, Retell around 85 to 90 and Synthflow around 75 to 80. The gap is real, but one strong year from Retell could close it.

If you want more recent data on this point, please see our latest agentic AI market report.

Which AI phone-agent startup handles the most real calls?

The call-volume race is now a three-company contest between Vapi, Synthflow and Retell.

Vapi reports between one million and five million calls per day. That converts to roughly 30 million to 150 million calls in a 30-day month. The range is wide, but even its lower end places Vapi among the largest platforms, while the upper end would make it more than twice Synthflow’s disclosed scale.

Synthflow says it handles more than 65 million calls each month, and Retell reports more than 50 million. Synthflow is therefore about 30% larger on their headline figures. Both numbers come from the companies themselves, so we cannot assume they use exactly the same definition of a call.

Bland handled more than 175 million calls during the previous year. Dividing that figure by 12 gives a monthly average of about 14.6 million, which places Bland at roughly 29% of Retell’s current volume and 22% of Synthflow’s. The calculation is imperfect because Bland’s activity may have accelerated during the year, but the order of magnitude is clear.

PolyAI says its agents operate across tens of millions of conversations, while Replicant has disclosed cumulative minutes rather than calls. Those figures prove real usage but cannot be normalized cleanly against the leaders.

Startup	Latest disclosed volume	Approximate monthly volume	How we treat it
Vapi	1M–5M calls per day	30M–150M	Company figure reported by TechCrunch
Synthflow	65M+ calls per month	65M+	Current company claim
Retell AI	50M+ calls per month	50M+	Company-reported figure
Bland	175M+ calls in one year	About 14.6M average	Derived from annual company figure
PolyAI	Tens of millions of conversations	Undisclosed	Directionally useful only
Replicant	1B cumulative phone minutes	Undisclosed	Minutes cannot be converted reliably

Which AI phone-agent startup is making the most money?

Retell has the strongest disclosed revenue story, with Parloa at a similar scale and Vapi probably somewhere in the same broad group.

Retell says it reached $50 million in annual recurring revenue. Parloa also passed $50 million during 2025, according to the Financial Times. Vapi has disclosed only a “healthy eight-figure” recurring-revenue run rate, which places it above $10 million but leaves too much room for an exact ranking.

Retell’s funding makes its result stand out. Dividing $50 million in reported recurring revenue by $5.1 million raised gives nearly $9.80 of ARR for each dollar of outside capital. Parloa’s equivalent ratio is about $0.09 because it has raised more than $560 million. Parloa raised much of that capital recently and has yet to spend it fully, but the difference is still enormous.

PolyAI’s latest available accounts showed £11.9 million in sales for the financial year ending in early 2025. That figure predates its latest financing and much of its recent expansion, so it probably understates the company’s current size. PolyAI nevertheless remains below the disclosed Retell and Parloa revenue level until fresher accounts prove otherwise.

Bland, Synthflow and Replicant publish convincing usage figures but no current revenue totals. Avoca has said it is on track to book $1 billion in customer jobs, although the value of jobs booked is customer revenue rather than Avoca’s own software revenue.

Retell wins on disclosed monetization and capital efficiency. Parloa may have a more durable enterprise contract base, but it has spent far more to reach a similar headline ARR.

Which AI phone-agent startup is growing fastest now?

Retell is currently growing fastest financially, while Synthflow has shown the sharpest jump in call activity.

Retell reportedly moved from about $5 million in ARR at the beginning of 2025 to $50 million by the end of that year. That is roughly tenfold growth in 12 months. As seen above, its current call volume has also passed 50 million per month, so the revenue expansion appears to be backed by real product usage rather than pricing changes alone.

Synthflow announced 45 million cumulative calls in mid-2025. Its current website now claims more than 65 million calls every month. The two numbers may have been counted slightly differently, but monthly activity now exceeds the entire cumulative figure it had previously announced. That is a major change in scale.

Replicant also appears to be recovering after a quieter period. It says revenue grew 2.5 times year over year following the launch of a new platform, with existing customers expanding automation to more types of calls. That growth comes from a more mature base than Retell or Synthflow, which makes it commercially valuable even if it is less dramatic.

Vapi’s current expansion is harder to calculate because it has not released a previous exact revenue baseline. Its rise in enterprise usage, large financing and growing customer list all point upward, but Retell provides the cleanest measurable growth curve.

If you want more recent data on this point, please see our latest agentic AI market report.

Who has won the strongest AI phone-agent customers?

Vapi owns the strongest single customer win, while PolyAI and Parloa have the broadest collections of major enterprise clients.

Amazon Ring gives Vapi unusually useful evidence. Ring evaluated more than 40 vendors, selected Vapi and moved all its inbound telephone traffic onto the platform. Ring also told TechCrunch that customer-satisfaction scores improved after deployment. A full rollout following a competitive process carries much more weight than a pilot or a logo on a customer page.

PolyAI now says it works with more than 200 enterprise customers across over 25 countries. Its named clients include Marriott, Caesars Entertainment, PG&E, UniCredit and Foot Locker. The company has also published measurable results, including a Forrester study estimating a 391% return on investment for a composite customer. The study was commissioned by PolyAI, so we should treat the exact percentage cautiously, but the underlying deployments are substantial.

Parloa’s customer list includes Allianz, Booking.com, HealthEquity, SAP, Sedgwick and Swiss Life. These are complex companies with long procurement cycles, sensitive customer data and multilingual operations. Winning several of them shows that Parloa can handle demanding enterprise sales and compliance reviews.

Replicant offers some of the clearest proof of expansion inside individual customers. DoorDash scaled its automation to tens of thousands of daily calls, while Sunrun reports that Replicant resolves half of its payment-related calls and receives a 4.6 out of 5 customer-satisfaction score.

Vapi wins on the quality of one competitive decision. PolyAI leads on breadth. Parloa looks strongest among heavily regulated global enterprises.

Which AI phone-agent platform is most mature and easiest to scale?

PolyAI has the deepest enterprise deployment history, while Vapi appears strongest at scaling raw infrastructure quickly.

PolyAI disclosed more than 2,000 live deployments across 45 languages and over 25 countries. Those deployments sit across banking, healthcare, hospitality, insurance, energy and retail. Repeating the process thousands of times requires mature implementation, monitoring and support teams, not only a strong underlying model.

Replicant has now automated more than one billion telephone minutes over roughly six years. Its customers include AAA, ADP, DoorDash, Fanatics and Sunrun. The company has therefore spent much longer dealing with production failures, call transfers, changing workflows and angry customers than most of the newer platforms.

Vapi has less history but greater recent infrastructure scale. Its enterprise product advertises 99.9% reliability and capacity for more than ten million calls, with dedicated support and service agreements. The company also says enterprise customers now account for most of its daily call traffic.

Retell and Synthflow have already crossed 50 million monthly calls, so both have moved beyond small deployments. Their shorter operating histories leave fewer public examples of how they perform through several years of customer expansion, model changes and unusual failure cases.

PolyAI leads maturity. Vapi leads elastic scale. Retell and Synthflow are closing the gap much faster than older enterprise vendors probably expected.

Which AI phone-agent startup has won developers?

Vapi has already won the developer layer of the AI phone-agent market.

More than one million developers have used Vapi, according to TechCrunch. The platform lets developers combine different speech-recognition models, language models, voice providers, telephone systems and business tools. Vapi says users can choose from more than 200 models and configure over 1,000 parts of an agent.

Retell is the nearest direct alternative. It offers Node.js and Python tools, webhooks, custom language models and integrations with services such as Twilio, Vonage, HubSpot, Make and n8n. It also includes a visual interface, which makes it more accessible to teams that mix developers with operations staff.

Bland provides an API and self-service product, but its current pitch increasingly centres on owning the full voice stack and handling high-stakes enterprise conversations. Synthflow has moved in the other direction by focusing on visual workflows and managed deployments.

Developer adoption creates a real advantage for Vapi. More users produce more integrations, examples, testing tools and engineers who already understand the product. A competitor can copy individual features much faster than it can reproduce that installed knowledge.

If you want more recent data on this point, please see our latest agentic AI market report.

Which AI phone-agent startup works best without engineers?

Synthflow is currently the strongest general-purpose choice for companies that want to launch phone agents without building the technical stack themselves.

Synthflow combines a visual builder, telephony, integrations, testing, analytics and implementation support. Its enterprise contracts now start at $30,000 annually, which shows that it has moved beyond a cheap do-it-yourself tool and is targeting larger deployments. As noted earlier, the company already claims more than 65 million calls per month across over 30 countries.

Parloa and PolyAI fit larger organizations that want the vendor to manage much of the implementation. Their projects tend to involve longer sales and deployment cycles, but they include governance, multilingual support and integration work that an internal team would otherwise have to handle.

Avoca is easier still for home-service companies because it arrives with more of the workflow already built. Its agents answer calls, book appointments, follow up on estimates and coordinate demand with technician capacity. Avoca says its customers are on track to book $1 billion in jobs through the platform this year.

A technical company that wants complete control should choose Vapi. A general business team should look first at Synthflow or Retell. A large international contact centre will probably prefer Parloa or PolyAI. For a heating, plumbing or roofing company, Avoca may deliver value faster than any horizontal platform.

Which startup has the best-performing AI phone agent?

No startup has proved that it has the best-performing AI phone agent today.

Vapi, Synthflow, Retell and Bland all advertise low response times, generally between about 400 and 600 milliseconds in favourable configurations. Those figures come from different tests using different voices, models, telephone carriers and regions. A platform can respond quickly and still misunderstand the caller or complete the wrong task.

The strongest independent evidence suggests that the whole category still struggles with complicated calls. The τ-Voice benchmark tested 278 realistic customer-service tasks. Voice agents completed between 31% and 51% under clean conditions, falling to 26% to 38% when the tests introduced noise, accents and more natural interruptions. The equivalent text agent completed 85%.

A newer benchmark published this summer reached a related conclusion. Voice systems performed very differently across naturalness, expressiveness, identity stability, emotion, accent and noise. A single score could not explain which system would work best for a particular customer.

Bland may have an advantage when customers need tight control because it owns more of the speech and language stack. Vapi gives developers the freedom to choose and tune individual components. PolyAI and Replicant have years of real customer conversations to improve their systems. Retell packages testing, quality assurance and fallback tools into a simpler product.

The buyer’s own calls remain the only convincing benchmark. A hospital appointment line, an insurance claim and an outbound sales call fail in completely different ways.

Which AI phone agent is cheapest at production scale?

Retell currently gives buyers the clearest all-in price, while Vapi offers the lowest visible platform fee for teams willing to assemble their own stack.

Vapi’s self-service plan charges for usage while passing model costs through to the customer. The headline platform price can look cheap, but the final bill also includes speech recognition, the language model, voice generation and telephony. HIPAA support costs an additional $2,000 per month on the standard plan, while zero data retention costs $1,000.

Retell publishes a range of $0.07 to $0.31 per minute. Its current calculator shows a representative configuration at $0.11 per minute, split between the language model, Retell’s infrastructure and the selected voice. Twenty concurrent calls are included on the self-service plan.

Bland advertises an all-inclusive price of roughly $0.11 to $0.14 per minute for its enterprise voice stack. Synthflow has moved to annual enterprise contracts starting at $30,000, with the final price depending on usage, telephony, integrations, security and implementation.

Platform	Public pricing	What buyers should understand
Vapi	Usage-based platform fee plus model costs	Lowest visible platform cost, but several components must be added
Retell AI	$0.07–$0.31 per minute	Clearest configurable all-in range
Bland	About $0.11–$0.14 per minute	More predictable bundled pricing
Synthflow	From $30,000 annually	Enterprise contract with implementation and support
Parloa, PolyAI, Replicant and Avoca	Custom	Requires a proposal and cannot be compared publicly

What can each AI phone-agent startup do that rivals cannot copy quickly?

Vapi has the broadest defensible advantage, while Bland, PolyAI and Avoca have narrower strengths that may be harder to reproduce inside their chosen markets.

Vapi’s advantage comes from its developer ecosystem and flexible orchestration layer. A rival can offer the same speech model or copy a dashboard feature, but recreating a community of more than one million developers, hundreds of integrations and years of testing across many configurations would take much longer.

Bland owns more of its speech, language and voice infrastructure. It supports self-hosted models, private deployments and enterprise compliance across areas such as HIPAA and PCI DSS. That gives Bland more control over cost, latency and data handling, especially for banks, insurers and healthcare companies.

PolyAI has accumulated operating knowledge from thousands of deployments and tens of millions of conversations. It also develops models specifically for customer-service dialogue. The value lies in the many unusual cases its systems have already encountered across languages, industries and countries.

Avoca’s advantage is vertical depth. By linking calls with bookings, technician schedules, estimates, marketing and customer records, it can learn which actions generate revenue for service businesses. A horizontal platform would need to build much of that workflow separately for every customer.

Retell currently wins through speed, simplicity and efficient execution. Those qualities have produced impressive growth, although they form a less obvious long-term barrier than Vapi’s ecosystem or Avoca’s vertical data.

If you want more recent data on this point, please see our latest agentic AI market report.

Which AI phone-agent startup has the strongest momentum lately?

Vapi has the strongest momentum as of now, while Bland and PolyAI have made the biggest recent moves behind it.

Vapi recently raised $50 million at a valuation of about $500 million, bringing total funding to $72 million. The company’s new capital arrived after large enterprise deployments, rapid call growth and the development of a wide self-service user base. That sequence is healthier than raising heavily before proving customer usage.

Bland raised another $50 million and now has more than 250 enterprise customers, hundreds of thousands of self-service users and over 175 million calls handled during the previous year. The company is using its proprietary infrastructure to target more complex financial-services, healthcare and insurance work.

PolyAI’s disclosed enterprise customer count rose from more than 100 in its late-2025 funding announcement to more than 200 in a company update a few months later. Definitions may have changed slightly, but doubling the published count so quickly suggests a real commercial push.

Parloa has the most financial firepower after raising $350 million at a $3 billion valuation. The new money should help it hire, expand internationally and compete for large accounts. For now, the financing has grown faster than the public evidence of deployments.

Vapi’s recent progress covers the widest range: funding, usage, developers and enterprise customers. Bland looks more formidable than it did a year ago, and PolyAI is expanding faster than its older reputation suggests.

Can we trust the AI phone-agent numbers?

The available data is good enough to rank the leaders, but the exact gaps remain uncertain.

Funding rounds and named customers are usually the easiest facts to verify. Revenue, call volume, latency and task-success rates often come directly from the startups. Private companies rarely publish audited breakdowns, and each company may count activity differently.

A reported “call” could mean a completed conversation, an attempted outbound call, a test call or another billable event. A company with many short calls can report more calls than a rival handling long and complex support conversations while processing fewer total minutes.

ARR also needs care. A recent TechCrunch investigation found that some AI startups use annualized run rates, expected contracts or unusually flexible definitions when discussing recurring revenue. We therefore treat Retell’s and Parloa’s $50 million figures as strong commercial evidence, without pretending they are identical to audited annual sales.

Vendor performance claims deserve even more caution. Latency measured in a quiet test will not reveal how the agent handles accents, background noise, interruptions, angry callers or a failed connection to the customer database. Independent benchmarks currently show a much weaker picture than most product demonstrations.

We are highly confident that Vapi, Retell and Synthflow form the leading phone-first group. We are also confident that PolyAI and Parloa lead parts of the managed-enterprise market. The precise distance between first, second and third could change once audited revenue, retention and comparable call data become available.

Which AI phone-agent startups are actually ahead?

Vapi is the overall AI phone-agent leader today, Retell is the closest challenger, and Synthflow completes a clear top three.

Vapi wins because it is strong across more important dimensions than any rival. It has enormous call volume, the largest disclosed developer base, growing enterprise revenue and the most convincing competitive customer selection. As seen above, Ring chose Vapi after reviewing more than 40 alternatives and then moved its full inbound workload onto the platform.

Retell ranks second by a fairly narrow margin. It has built a comparable order of commercial scale using a fraction of the funding raised by most competitors. Its biggest missing pieces are independently reported customer deployments, detailed retention data and evidence that its current growth can survive several years of enterprise use.

Synthflow takes third. Its call volume is already higher than Retell’s disclosed figure, and its no-code approach opens the market to companies without dedicated voice engineers. We keep it below Retell because its revenue, customer depth and retention remain less visible.

Parloa ranks fourth as the strongest heavily funded enterprise contender. PolyAI follows closely because it has more deployment history and a broader disclosed customer base. Parloa’s growth and capital place it slightly ahead today, while PolyAI could move above it once newer financial results appear.

Bland comes sixth, although it may rank much higher for regulated or private deployments. Replicant remains a proven enterprise company with improving growth, but it has lost some visibility to faster-moving rivals. Avoca ranks last only in the broad horizontal comparison. It is already one of the strongest choices inside home services.

The ranking could change in three clear ways. Retell could take first by maintaining its growth and publishing deeper enterprise proof. Synthflow could move above Retell by showing that its call volume produces comparable revenue and renewals. Parloa could enter the top three if its huge financing translates into clearly larger deployments rather than mainly a larger organization.

Rank	Startup	Why it ranks here
1	Vapi	Best overall mix of call scale, developers, enterprise validation, flexibility and recent momentum
2	Retell AI	Exceptional revenue growth and capital efficiency, with usage already close to the leader
3	Synthflow	Major call volume and the strongest no-code position, but less revenue and customer transparency
4	Parloa	Strong enterprise revenue, blue-chip customers and unmatched financial capacity
5	PolyAI	Deepest deployment history and broad international customer base
6	Bland	Proprietary infrastructure, strong regulated-industry positioning and rising enterprise adoption
7	Replicant	Proven production history and renewed growth, but less current momentum than the leaders
8	Avoca	Powerful vertical product for home services, with less relevance outside that niche

If you want more recent data on this point, please see our latest agentic AI market report.

Who is the author of this content?

NEW MARKET PITCH TEAM

We track new markets so founders and investors can move faster

We build living "market pitch" documents for emerging markets: AI, synthetic biology, new proteins, and more. Instead of outdated PDFs or hallucinated LLM answers, our clients get a clean, visual, always-updated view of what's really happening: key players, deals, regulations, and signals that matter. Learn more about us.