ElevenLabs goes from voice tool to enterprise agent platform with 13 new products
ElevenLabs announced 13 new products and upgrades at its New York Summit on November 11, moving from voice generation into enterprise conversational AI — including ElevenAI, a workplace app that turns meetings into notes and action items, and an expanded Nvidia deal putting Nemotron 3.5 Super VL into ElevenAgents. About half the package is preview, early access or roadmap rather than generally available, and ElevenLabs' claim that Nemotron completes support tasks at 70% lower cost is its own benchmark, not independently verified.
📊 Scorecard: our Bullish call on $NVDA is scored from $173.73, the last close before it ran. Its first result comes after the next close.
The News
ElevenLabs announced 13 new products and upgrades at its New York Summit on November 11, in a deliberate move from voice generation into enterprise conversational AI. The headline product is ElevenAI, a workplace app that listens to meetings — including in-person ones — and turns them into notes, summaries and action items, with free early access opened on November 11. ElevenLabs is a private company, so the public read-through is Nvidia: the Summit expanded its Nvidia partnership, putting the Nemotron 3.5 Super VL model into ElevenAgents and planning 15 megawatts of Vera Rubin compute capacity on top of its existing Blackwell Ultra systems.
The rest of the package fills out the platform. Model customisation will let selected enterprises work directly with ElevenLabs researchers to fine-tune open-source language models and ElevenLabs voice models on live deployment data, starting in the coming months, with no pricing announced. ElevenAgents — the company's agent builder — gains Memory, so agents remember past conversations; Routines, for repeating workflows; and Architect, for designing agents; plus a previewed Interaction Engine that picks the right model in the middle of a conversation. A Partner Program of more than 550 partners across 44 countries, and a Marketplace of partner-built industry apps, give the platform a sales channel.
Two caveats belong in the open. First, about half the package is preview, early access or roadmap rather than generally available — ElevenAI's free early access is live, but the Interaction Engine is a preview and model customisation has not started yet, so the immediate effect is a platform pitch to enterprise buyers, not a finished catalogue. Second, ElevenLabs says Nemotron completed its customer-support benchmark tasks at 70% lower cost than other leading open models — its own benchmark, not independently verified.
ElevenLabs is no longer selling a voice — it is selling the agent that speaks with it.
Why It Matters
ElevenLabs' move makes sense because voice is becoming the front door to AI agents. A text chatbot is useful; an agent you can talk to — that takes a support call, listens in a meeting, or speaks to a customer — is a product companies will pay for. ElevenLabs started with voice generation, the ability to clone and synthesise speech, and it is now building the rest of the stack around it: the agent, the memory, the marketplace. The Summit was the moment the company stopped being a voice vendor and started being an agent platform.
The Nvidia deal is the most concrete part of the announcement. Nemotron 3.5 Super VL is Nvidia's own open vision-language model, and putting it inside ElevenAgents — plus 15 megawatts of Vera Rubin capacity, Nvidia's next-generation AI chips — ties ElevenLabs' platform to Nvidia's hardware roadmap. For Nvidia, every agent platform that runs on its models and its chips is another demand channel; for ElevenLabs, the deal is a stamp of credibility with enterprise buyers who want to know the platform will scale.
But the 70%-lower-cost claim needs the label it carries: it is ElevenLabs' own benchmark, not an independent test, and 'other leading open models' is a vague comparison set. Cost claims from a vendor about its own partner's model should be treated as marketing until an independent lab reproduces them. The same caution applies to the package as a whole: a Summit announcement is a pitch, and roughly half of what was pitched is not yet available to buy.
Analyst's Take
Our call is Bullish. ElevenLabs has a real product in ElevenAI, a real platform story in Memory, Routines and Architect, and a real Nvidia deal behind it — 15 megawatts of Vera Rubin capacity is not a press-release number, it is a build-out. The company is moving from a single-feature voice tool into the enterprise agent market at exactly the moment that market is forming.
What to watch: whether the preview and roadmap items — the Interaction Engine, model customisation — actually ship, and whether independent tests back the 70% cost claim for Nemotron. A platform is judged by what customers can buy, not by what a Summit announces. If the previews convert to products, ElevenLabs has made its move; if they linger, the Summit was a pitch.
Are you bullish or bearish on this story?
By The Numbers
FAQ
What is ElevenLabs?
ElevenLabs is a company best known for AI voice generation — realistic synthetic speech. At its New York Summit on November 11 it announced a move into enterprise conversational AI, building voice agents for businesses.
What is ElevenAI?
ElevenAI is a workplace app that listens to meetings — including in-person ones — and turns them into notes, summaries and action items. Free early access opened on November 11.
What is Nemotron 3.5 Super VL?
It is Nvidia's open vision-language model — an AI that can understand both images and text. ElevenLabs has added it to ElevenAgents, its agent builder, as part of an expanded deal with Nvidia.
What does 15 MW of Vera Rubin capacity mean?
Vera Rubin is Nvidia's next generation of AI chips, and 15 megawatts is the amount of computing power ElevenLabs plans to run on them — a measure of how much AI compute the company is building out, on top of its existing Blackwell Ultra systems.
Is the 70% lower cost claim proven?
No. It is ElevenLabs' own benchmark, not independently verified, and the comparison is against 'other leading open models' without naming them. Treat it as the company's claim until an independent test reproduces it.
👁
Source: CIOL