Provides strategic insight into the cost and adoption trajectory of voice AI for customer service operations.
What is ElevenLabs’ revenue and valuation?
ElevenLabs builds the voice layer of AI, the models that turn text into speech that sounds human. TechCrunch reports the company is pacing at 600 million dollars in annual recurring revenue, with backers valuing the 4-year-old company at 22 billion dollars.
Classic enterprise accounts generate 55% of that revenue, and the remaining 45% comes from small and medium businesses, developers, builders, and creators. The split matters for buyers, because a platform carried by enterprise contracts has the runway to keep improving the models you depend on. Staniszewski co-founded the company and still draws the model line himself, which is a governance signal worth checking for at any voice vendor: the person pricing the platform is the person deciding where the stronger models get used.
ElevenLabs runs at 600 million dollars in ARR with enterprise buyers carrying 55% of it.
Which companies run customer service on ElevenLabs?
Klarna runs first-line phone support for 35 million U.S. customers on the platform. Deutsche Telekom, Cisco, Adobe, and a growing list of governments run on it too, which is the clearest signal that the technology cleared the production bar. The customer list now includes competitors in the making, with conversational AI vendor Decagon having trained its voice product on ElevenLabs and now running queries through its own models, a shift the CEO acknowledged in the interview.
Government deployments show the range. Poland’s public health system uses ElevenLabs agents to call patients who book appointments, chasing an 18% no-show rate with reminders while keeping data residency inside the country. Creators use the same underlying platform for audiobooks, dubbing, and music, which is why the model quality keeps climbing across use cases.
Millions of live customer conversations run on ElevenLabs infrastructure today.
When should voice AI use frontier models instead of open source?
CEO Mati Staniszewski drew the line himself in the TechCrunch interview. Informational queries, where the knowledge base defines a good experience, run well on open-source weights. Financial services calls that need authentication, transaction details, or refunds have no room for error, and frontier models lead there. His stated ambition is passing the Turing test for conversational AI, an agent that reads the caller’s emotions and slows down or speaks up in response, which tells you where the quality bar is heading.
The same logic maps onto any small business queue. The calls that ask when the store opens are cheap to automate, and the calls that move money or touch compliance deserve the stronger model. Matching the model tier to the call type is the difference between a pilot that saves money and one that creates incidents.
Match the model tier to the call: informational queries on open weights, money and identity on frontier models.
What does ElevenLabs’ growth mean for support budgets?
The 600 million dollar pace means enterprise buyers already trust voice agents with real customer conversations. A support line that routes repetitive calls to an agent stops paying humans to read scripts, and the cost case lands on the highest-volume tier first.
For teams evaluating the category, our ElevenLabs intelligence report covers the platform’s pricing and deployment options, the vendor’s own documentation is the source of record for current limits, and the adoption curve is visible in enterprise partnerships like the DXC rollout announced in July, which embeds ElevenLabs voice into customer service platforms at scale.
Support budgets are the line item this growth story is aimed at.
The after-hours line at a property management company rings into a voicemail box, and by morning the box is full. Every missed call is a tenant with a leak, a locked-out owner, or a lease question calling the next company on the list.
A voice agent that answers at 2 a.m., books the emergency, and escalates the rest works like a night desk that never sleeps. The same math that moved Klarna’s 35 million U.S. support calls onto ElevenLabs applies to a 12-person firm: the volume is smaller and the payoff on each call is bigger.
That is what 600 million dollars of ARR represents. Buyers at every size already trust these agents with the calls that used to sit in voicemail.
Should your business deploy voice AI agents now?
Pilot it on the informational tier now and keep authentication and payments on stronger models. The CEO’s own disclosure advice is worth adopting from day 1: tell callers they are talking to an agent, and when the wait for a human is 30 minutes, offer the choice, because customers pick the agent and are surprised by the quality.
Evaluate your current support volume to find the repetitive inbound calls that waste staff hours, and route those first. Vendor pricing moves as competition intensifies, so a review of the cost per resolved call each quarter keeps the math honest. Every ElevenLabs customer passes KYC checks and the platform blocks agents that create more agents, per the CEO, which are the guardrails worth demanding from any voice vendor you sign.
Route your repetitive informational calls to a voice agent this quarter, with disclosure on the line.
Source: TechCrunch AI
