AI agents are no longer just experiments — and ElevenLabs’ latest numbers make that clear.
The company shared that enterprises and developers built more than 3.3 million agents on its platform in 2025 alone, just over a year after launching ElevenLabs Agents. It’s a big jump, and it shows how fast businesses are moving from “testing AI” to actually using it in daily workflows.
Alongside the milestone, ElevenLabs also looked back at what changed on the platform this year — and most of it was about making agents less fragile and more useful in the real world.
One of the most noticeable upgrades was automatic language detection. Agents can now recognize what language someone is speaking and switch on the fly, without needing manual setup. For global teams, that’s the difference between a demo and a usable product.
Another key change was a new turn-taking model, which helps agents understand when to speak and when to stay quiet. It’s a small thing, but it makes conversations feel far less awkward — especially in voice interactions.
As more companies started deploying agents at scale, ElevenLabs introduced Agent Workflows, giving teams clearer control over how conversations move from one step to the next. This made agents more predictable and easier to trust in production environments.
Before launch, teams can now test agents using real-world conversation simulations, checking behaviour, safety, and compliance issues early instead of discovering problems later with customers.
The platform also expanded beyond voice. Chat Mode brought agents into text-based experiences, while WhatsApp support opened access to the world’s most widely used messaging app. For businesses that rely on phone systems, IVR navigation support made it possible for agents to handle routine call flows.
On the infrastructure side, ElevenLabs rolled out versioning, so teams can track changes and roll out updates safely. A new speech-to-text model now powers the Agents Platform as well, improving accuracy across conversations.
Taken together, these updates explain why usage has grown so quickly. ElevenLabs isn’t just adding features — it’s trying to make AI agents stable enough to rely on.
And judging by the 3.3 million agents built this year, a lot of teams are already doing exactly that.
