ElevenLabs bets the company on enterprise agents: 13 products, a meeting app, and a bigger Nvidia deal
ElevenLabs spent four years becoming the default name in AI voice. On November 11, at its first New York Summit, the company made its case for becoming something bigger: the layer enterprises build their conversational AI on. Thirteen products and upgrades, a new workplace app, custom models tuned on each customer’s own data, and a deeper Nvidia partnership — all announced in one keynote, and all pointed at the same buyer.
- Announced November 11, 2026 at ElevenLabs’ New York Summit: 13 new products and upgrades aimed at enterprise buyers.
- ElevenAI: a new workplace app that turns spoken conversations, including in-person meetings, into notes, summaries, and action items. Free early access is open now.
- Custom models: selected enterprises will fine-tune audio models and LLMs on their own live deployment data, refined with reinforcement learning. Work begins in the coming months; pricing is undisclosed.
- Nvidia deal: Nemotron 3.5 Super VL is now available in ElevenAgents, and ElevenLabs plans to add 15 MW of Vera Rubin compute capacity.
- Caveat: about half the package is preview, early access, or roadmap rather than generally available — and the 70% lower-cost claim for Nemotron is the company’s own benchmark, not independently verified.
ElevenLabs spent four years becoming the default name in AI voice. On November 11, at its first New York Summit, the company made its case for becoming something bigger: the layer enterprises build their conversational AI on. Thirteen products and upgrades, a new workplace app, custom models tuned on each customer's own data, and a deeper Nvidia partnership — all announced in one keynote, and all pointed at the same buyer.
The pitch, in CEO Mati Staniszewski's words, is that conversational agents will become "critical infrastructure for every enterprise, as essential as cloud or payments." The announcements are the company's attempt to sell the infrastructure, not just the voice.
From voice tool to enterprise platform#
The package falls into four parts: models, agents, a workplace app, and a partner channel. The through-line is customisation. Where ElevenLabs once sold a voice generator anyone could use the same way, it now wants to sell systems shaped to each business — its data, its customers, its workflows.
Three new voice models arrived in the month before the summit: Scribe v3 for speech-to-text transcription, Eleven v4 for expressive text-to-speech, and Eleven Realtime, a speech-to-speech model built for fast voice agents. The company says all three lead independent benchmarks. It also launched Scribe v3 Medical, a transcription model tuned for healthcare, and said more industry-specific models — financial services is next in line — are on the way.
The bigger model news is what comes next. Selected enterprises will work directly with ElevenLabs researchers to fine-tune audio models and language models on data from their own live deployments, then keep refining them with reinforcement learning against metrics like customer satisfaction and reliability. "We believe in a future where each of you will have your own RL stack," Staniszewski said on stage. The work starts in the coming months, for selected enterprises first, and ElevenLabs has not disclosed pricing — so whether this stays a large-enterprise offering is an open question.
Agents that remember — and act#
ElevenAgents, the company's platform for building voice and chat agents, gets three upgrades aimed at the gap between a good demo and a useful employee. Memory gives agents the full context of returning customers, so nobody has to repeat themselves. Routines run multi-step work before, during, and after every conversation — processing leads and follow-up actions without waiting to be asked. And Architect, a built-in expert, helps teams build and improve agents by talking or typing, drafting each proposed change, testing it in simulation, and holding it for approval.
Alongside them sits the Interaction Engine, a system for orchestrating voice models and language models across a single conversation. The premise is that not every step of a task needs a frontier model: the engine picks the right model for each part, balancing performance and reliability against token cost. "Interaction Engine orchestrates all this under the hood," Staniszewski said, "so you can be confident you're delivering the right outcome for the customer, without wasting money on tokens."
ElevenAI listens to the meeting#
The most immediately usable announcement is ElevenAI, a new app that acts as an AI teammate for enterprise teams. It listens to spoken conversations — including in-person meetings — and turns them into notes, summaries, and action items, with notice and consent mechanisms built in. It can also offer live assistance during a call. Free early access opened with the announcement, so anyone can try the meeting-notes experience now rather than waiting for a wider rollout.
It puts ElevenLabs in direct competition with the crowded meeting-notes field, where the differentiator is supposed to be the voice stack underneath: transcription, understanding, and spoken assistance from one vendor instead of three.
The Nvidia thread#
The infrastructure behind the push comes from an expanded Nvidia partnership. Nvidia's Nemotron 3.5 Super VL, an open vision-language model built for reasoning in agentic workloads, is now available inside ElevenAgents. ElevenLabs says it completed customer-support benchmark tasks at 70% lower cost than other leading open models — its own benchmark, not independently verified, so treat the number as a claim rather than a result.
The hardware commitment is larger still: ElevenLabs plans to add 15 megawatts of Vera Rubin compute capacity on top of its existing Blackwell Ultra systems. The two companies are also working on an on-premises voice agent for Centralna e-Rejestracja, the appointment system for Poland's national e-Health platform, delivered with EasyCall — a sovereign deployment, and a calling card for regulated industries wary of sending voice data to US clouds.
To reach those buyers, ElevenLabs launched a Partner Program counting more than 550 partners in 44 countries, an alliance with McKinsey to help organisations build and scale agents, and a Marketplace where companies can adopt agents already customised for healthcare, financial services, insurance, education, and retail by more than 20 partners. On stage, Yahoo showed a voice-powered AI assistant and Westgate a new front-desk agent; Stripe, TELUS, FOX Sports, Klarna, and SevenRooms by DoorDash already run ElevenAgents in production.
What to watch#
Two caveats belong in the open. First, about half of 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. The immediate effect is a platform pitch to enterprise buyers, not a finished catalogue. Second, the headline numbers — benchmark leads, the 70% cost figure — are the company's own, and pricing for the custom-model work is undisclosed.
What matters now is execution: whether Memory and Routines survive contact with real support queues, whether the McKinsey alliance turns into signed enterprise deals, and whether fine-tuning on live deployment data becomes a product mid-sized companies can buy — or stays a concierge service for the largest ones. ElevenLabs is no longer selling a voice. It is selling the agent that speaks with it, and the infrastructure underneath.
Sources: Mumbrella — ElevenLabs launches 13 new products and upgrades (Nov. 12, 2026); AI Agent Update — ElevenLabs goes from voice tool to enterprise agent platform (Nov. 12, 2026); TechInKey — ElevenLabs ElevenAI: Custom Models and NVIDIA Deal at New York Summit (Nov. 12, 2026); FreshSynapse — ElevenLabs Unveils Its Interaction Engine and 13 Upgrades (Nov. 12, 2026).