# ElevenLabs Previews Interaction Engine, Announces Upcoming Customization Tools

By Simon Yoon

Canonical URL: https://www.tokenpost.com/news/technology/55117
Published: 2026-11-12T08:11:47.000Z
Updated: 2026-11-12T08:11:47.000Z
Section: Technology

> The system is designed to manage turn-taking, interruptions and model selection; businesses will soon be able to fine-tune language and voice models with their own data.

ElevenLabs previewed an engine for managing AI conversations and outlined future model-customization tools at its Nov. 11 New York summit, giving businesses new options for building voice and chat agents.

The previewed system is designed to coordinate the back-and-forth of a voice or chat exchange. It manages turn-taking and interruptions, then selects a language model suited to the request. Simpler tasks can be sent to lower-cost models, while more complex interactions can be assigned to advanced models. ElevenLabs described this routing approach as a way to optimize performance and reduce operating costs. The system is intended to avoid using the most advanced model for every request by adjusting model choice to task complexity.

Separately, ElevenLabs announced that businesses will soon be able to fine-tune open-source language models and the company’s voice models with their own data. The customization tools are intended to help customers optimize for measures such as customer satisfaction scores. Businesses will also be able to create reinforcement-learning stacks tailored to their needs.

The company demonstrated Scribe v3 Medical, a speech-to-text model built for health care terminology. It can distinguish between similar drug names, including cycloserine and cyclosporine, a capability intended to reduce errors in specialized contexts. ElevenLabs also plans to develop a model for financial services.

Updates to ElevenAgents, the company’s platform for deploying and managing business voice and chat agents, include Memory. The feature lets an agent retain details shared in an earlier interaction and use them in later conversations. Examples include a customer’s language preference or an availability constraint. Memory is intended to help an agent build a cumulative understanding of each customer and personalize future interactions.

Routines can automate follow-up work after a customer conversation. ElevenLabs designed the feature to streamline workflows and reduce the manual effort involved in post-conversation activities.

CEO Mati Staniszewski described conversational AI as a technology likely to become as essential for enterprises as cloud computing or payment systems. “Every business will need to reinvent how it interacts with its customers, employees, and the wider world,” he said.