ElevenLabs' Tripled Agent Revenue Makes Voice a Front-Line GTM Channel
ElevenLabs closed a $300 million tender at the $22 billion valuation first reported in July. The new number is agent revenue, up threefold since February, from companies putting AI on their phone lines.
October 1, 2026
ElevenLabs said on September 30 that it had closed a $300 million employee tender at a $22 billion valuation, led by Wellington and T. Rowe Price. In a tender, investors buy existing shares from employees. ElevenLabs paired the news with numbers it hadn't disclosed before. Revenue from ElevenAgents, its platform for voice and chat agents, has more than tripled since February. The agents now handle more than 15 million conversations a week, also up threefold. Enterprise customers account for 55% of revenue.
The valuation was already known. Bloomberg reported in early July that ElevenLabs was in talks for a tender at about $22 billion, as The Next Web summarized at the time. That was double the $11 billion it set with a $500 million Series D in February. On September 30, the price held through the close. The company told buyers where the growth behind it comes from. TechCrunch's report on the close leads with the doubled valuation. For a revenue team, the agent numbers matter more.
A week earlier, CEO Mati Staniszewski told TechCrunch the company was at $600 million in annual recurring revenue, with more than half from enterprise accounts. The first customer Staniszewski named was Klarna, which runs first-line phone support for 35 million US customers on ElevenLabs. Deutsche Telekom, Cisco and Stripe are also named customers. A company best known for cloned voices in audiobooks and dubbing now earns a growing share of its money answering phones for other companies.
Voice agents now compete with a staffed minute
That shift changes who the buyer is and what the product gets compared to. Creators bought ElevenLabs as a production tool and weighed it against hiring a voice actor. Enterprises buy ElevenAgents to take calls, so the obvious comparison is a staffed minute. In 2025, ElevenLabs cut its agent pricing to $0.10 a minute on its smaller plans and $0.08 on annual business plans, with model costs billed on top. A fully loaded support rep or inbound qualifier costs many times that per working minute, and a CFO can do that math quickly. With tripled agent revenue and 15 million weekly conversations on record, finance can ask for that comparison by name.
This is where voice enters customer-facing GTM. Staniszewski's LinkedIn post announcing the close lists sales alongside support and operations as the work the agents do. The obvious use is tier-one support, and that's where Klarna sits. But most of the calls that look like support from the outside are mostly revenue motions. They include inbound lead qualification, appointment and renewal reminders, payment follow-ups, and the first call after a demo request. An SDR team or a renewals desk staffs those calls today, and they're scripted enough for an agent to run. Staniszewski also told TechCrunch that callers should be told when they're talking to an AI. That matters for anyone putting an agent in front of a prospect, because a disclosed bot on a qualification call is a brand decision first and a cost decision second.
Klarna and Apple kept humans behind the agent
Klarna is also the best evidence against the agent case. In 2025 Klarna’s CEO said its AI-first support push had cost it quality. The company said it would hire human agents again so customers could always reach a person. Apple offers a more recent example. Bloomberg reported on September 29 that Apple drafted a plan over the summer to replace about 5,000 AppleCare workers with AI agents and then shelved it, per The Next Web's account of the report. Apple still routes its US and Canada support line through a generative AI assistant. Both companies kept the agent at the front door and kept humans behind it. The voice layer is winning the first minute of the call. Claiming that much is safer than claiming voice agents will replace call centers.
The ElevenLabs numbers have limits too. The tripling is a ratio from the company's own announcement. It gives no absolute figure for agent revenue, so there's no way to tell how much of the $600 million it represents. Competition is also closing in. Staniszewski acknowledged that Decagon, which built its voice product on ElevenLabs, now competes with it directly. Public reaction so far is funding recaps and builder demos on LinkedIn, X and Reddit. We found no operator accounts of an agent replacing a staffed sales or renewals line.
The signal still holds. Two years ago this company sold a creative tool. Now it’s valued partly on the volume of business calls it handles, and enterprises are buying that volume at cents per minute. A CFO can compare that with headcount using the public numbers alone.
Since this piece ran, an operator account has surfaced, though of an agent working in front of a sales team rather than replacing one. On October 19, monday.com launched Harmony, voice agents for mid-market and enterprise sales teams, and its only named launch customer, a UK consumer car-finance broker, says the agent now follows up with every approved applicant right away, qualifies about half of those conversations and hands 80% of them live to its advisers. That's inbound qualification run as an immediate outbound follow-up, one of the motions this piece suggests piloting first, and the broker credits it with a 9% increase in deals closed in less than a year, a figure it reports with no holdout, from consumer lending rather than a B2B sales team. Harmony's public rate card starts at $0.39 a minute of agent time plus a $250 monthly platform fee on its Pro plan, a price to set against ElevenLabs' rates and the cheaper voice models Microsoft released on October 1. We looked at what Harmony's per-minute pricing means for sales buyers in a separate piece.
What this means for CROs and customer success leaders
If you are a CRO or a customer success leader, you should get ahead of that comparison with a pilot you control. Pick one narrow, scripted motion, such as inbound qualification for small accounts or renewal reminders 60 days out. Track resolution and handoff rates, and watch what happens to conversion when a human picks up after the agent. Tell callers they’re talking to an agent. Then take the per-minute math to finance yourself, including the handoff cost, before it shows up as a vendor's slide in a budget review.
My bet is that the inbound qualifier gets benchmarked against a voice agent before the support rep does. Support already has its Klarna and Apple stories. Inbound qualification has a cleaner metric, meetings booked, which makes it easier for finance to price.
Both companies kept the agent at the front door and kept humans behind it. The voice layer is winning the first minute of the call.
Pilot one narrow, scripted motion, such as inbound qualification for small accounts, track handoff rates, and take the per-minute math to finance before it shows up on a vendor's slide.
