Retell Reports a $100M Voice AI Run Rate on a Single $4.6M Seed Round

Retell, which builds AI voice agents for contact centers, says it passed $100 million in annualized run rate on $4.6 million of total funding.

The company said on November 12 that the figure grew from $10 million 17 months earlier, and it's launching Tell-1, a model of its own built for live calls. Retell says it has been profitable since 2025 with fewer than 80 employees, on a single seed round led by Alt Capital. Retell bills by usage with no minimums, so the figure is annualized usage rather than contracted ARR.

Retell credits a motion that lets anyone build an agent without a sales call, and its fastest growth now comes from enterprise contact centers.

Why phone calls are hard to automate

The phone is still how patients, policyholders and borrowers reach healthcare, insurance and financial companies. Retell calls it the hardest channel in customer experience to automate. On a live call, the company says, half a second of lag or one mistimed interruption can turn a resolved problem into a hang-up.

CEO Bing Wu told Forbes the modern contact center "is built around major constraints" in how many calls it can take and what its agents remember. Callers end up repeating their name and Social Security number. "It's not going to be the old world plus AI on top," Wu said.

Cost is the other pull. Sacra, which sells research on private companies, estimates Retell's per-minute price runs 70% to 95% below offshore human agents, who cost $0.30 to $0.80 a minute.

Where Retell sits in voice AI

Retell sells the layer that runs a call. Its software decides when the agent speaks, listens or handles an interruption, according to OpenAI's customer story on Retell. It also calls the customer's own systems and briefs a human before handing the caller over. Wu calls this voice agent orchestration, which he defined to Forbes as "how you design and orchestrate the voice experience."

ZK Research analyst Zeus Kerravala told Forbes the market splits in two. One side is contact center platforms such as NICE, Talkdesk and Amazon Connect. The other is conversational AI specialists such as Kore.ai, PolyAI and Omilia.

Sacra adds speech-model makers such as ElevenLabs and Deepgram building their own agent platforms, and support-agent companies such as Sierra and Decagon moving from chat into voice. The closest match to Retell is Vapi, which also started with a self-serve platform for developers. Vapi has raised $72 million, has about 100 employees and runs at an eight-figure run rate, an investor told TechCrunch in May.

Retell stays model-agnostic against all of them. OpenAI models generate responses on 80% of its calls. Retell ranks the models it offers on its own public benchmark, built from real production conversations.

Retell went from $10 million to $100 million in 17 months

Retell was founded in 2023 by Wu, CTO Zexia Zhang, CPO Evie Wang, Weijia Yu and Todd Li, who had worked at Meta, TikTok, Google and ByteDance. It started as a live dubbing tool for creators before turning to phone agents, according to Sacra. Jack Altman, whose Alt Capital led the seed, told Forbes that in spring 2024 "it just seemed to me that voice was becoming inevitable."

Retell's About page dates its milestones:

• May 2024: a $4.6 million seed, its only round, three months after launch.
• June 2025: $10 million in annualized run rate, 13 months after the seed.
• November 2025: $40 million in annualized revenue, reached with a team of 25 while fully profitable.
• September 2026: $80 million.
• November 2026: $100 million.

Monthly calls on the platform grew from 30 million in early 2026 to more than 100 million by September, according to OpenAI's story. Retell reports $1.6 million in annual run rate per employee. Sacra, the one outside source to size the business, estimated about $60 million in annualized revenue in April 2026, up 650% from a year earlier.

Decagon took far more capital to reach the same mark. It told Newcomer in August that its annualized revenue passed $100 million after raising about $481 million, according to Value Add VC.

How Retell sells

"Every company with a phone line has been told AI will transform its contact center, and most have been shown a scripted demo," Wu said in the release. Retell's answer is to let anyone sign up and build an agent without talking to sales, and to give every customer the same product.

That started with smaller buyers. In May 2024, Retell told TechCrunch it had hundreds of customers, all paying per minute, including telehealth company Ro. Sacra says Retell found its fit with outsourced call centers such as Everise, lender Sunshine Loans and insurance company Matic.

Larger customers get forward-deployed engineers who help them launch. Wu describes the end state as customers building "an in-house team of internally deployed engineers who own and run it all," per OpenAI's story.

Retell also sells through partners. Its certified partner program, which Wu announced in November 2024, covers app builders, technology integrations and resellers, with co-marketing and profit-sharing.

When Retell launched automated quality assurance in December 2025, Wu said it was the top request of its enterprise customers, "our fastest-growing customer segment." Retell says enterprise revenue rose ninefold in 2026, led by healthcare, financial services, insurance and legal.

Kerravala told Forbes those buyers historically signed "multi-year, multi-million-dollar platform contracts." In Kerravala's view, Retell's rise shows enterprise teams want "transparent consumption pricing and direct ownership of their agent workflows."

Customers and the results they report

Retell names Lenovo, BILL and CVS Health among its customers. The release and OpenAI's story list these results:

• Asbury Automotive Group saves $2 million a year, and its appointment-set rate rose 40%.
• Anker resolves 80.4% of its tier-one post-sales support cases with Retell agents, with a Net Promoter Score of 63.
• Plymouth Rock automates 75% to 80% of its support calls.
• Morgan & Morgan estimates its cost per minute is 55% lower than with human agents.
• Pine Park Health freed medical-assistant time equal to 2.1 full-time roles, and its scheduling agent books 55.7% of the patients it reaches.
• The San Antonio Spurs say 14% of calls end in a payment agreement.
• Sunrun answers all of its inbound transfer calls with Retell first and runs 22 agents built by its own team.

Retell's next products are built for its enterprise buyers

Tell-1 was trained on hundreds of millions of calls and responds in under 100 milliseconds, Forbes reported. "Latency is the foundation for everything," Wu said. Wu added that "in order to build the best application, you have to own the model as well."

Tell-1 gives Retell a model of its own next to the OpenAI models that answer most of its calls. Retell is also launching Retell Dialer, which runs outbound campaigns across voice and SMS agents, such as calling a prospect back at the time they asked for.

Retell plans to expand quality assurance and simulation next, support more languages and build knowledge bases that update themselves, according to OpenAI's story. The quality assurance work extends the feature enterprise customers asked for most in 2025.

Enterprise now sets the pace for a product anyone can sign up for

Retell's fastest growth comes from customers that launch with its engineers beside them, in healthcare, financial services, insurance and legal. Those are the buyers Kerravala says once committed to long platform contracts with vendors such as NICE and Talkdesk.

If enterprise keeps outgrowing self-serve, more of Retell's revenue likely depends on its engineers at launch. Wu's stated goal is the reverse, with customers running the agents themselves. Sunrun already runs 22 agents its own team built.

Every company with a phone line has been told AI will transform its contact center, and most have been shown a scripted demo

Retell plans automated QA and simulation, more languages and self-updating knowledge bases next, built for enterprise customers whose revenue grew ninefold in 2026.

By Business of GTM