Zendeskannounceda new Real-Time Voice Translation capability on Thursday. It provides bidirectional translation during inbound and outbound contact center voice calls. The customer and agent can each speak and hear the conversation in their preferred language, without involving a human interpreter, moving to another channel or using a separate desktop application.
The translation capability is built into the native Zendesk Contact Center experience. A customer’s language can be identified in the contact flow or selected by the agent, who can control translation without leaving the call. Administrators can enable it by queue and choose whether recordings retain the original audio, translated audio or both.
Early access program for Real-Time Voice Translation
A closed early access program (EAP) Real-Time Voice Translation begins in October for eligible Zendesk Contact Center customers, with general availability planned for Q1 2027. Customers will not be charged during the EAP, although usage will be capped.
During the EAP, the offering will support Chinese, English, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Portuguese, Russian, Spanish and Vietnamese. Zendesk expects that to expand after the EAP based on customer requirements.
The EAP version of the service will have some limitations: It will not translate multiparty or transferred calls, supervisor monitoring or barge-in or video calls. It also lacks viewable usage reporting and administrative usage limits.
Zendesk has not settled on a commercial model for general availability. “It will be usage-based, but we want to better understand usage in the early access program before defining what the price is going to be,” said Lilith Shoemaker, product marketing manager, Zendesk.
AI broadens translation use
The announcement brought me back to a post I wrote in 2013,“Video in the Contact Center: Just Do It.”LanguageLine had recently introduced LanguageUc, an app-based video remote interpreting service, as a promising answer to contact center language barriers. By 2015, video accounted for just 3% of LanguageLine’s revenue, while telephone interpreting accounted for 86%.
Until recent advances in AI, real-time translation remained an expensive and cumbersome service for contact centers to offer, restricted to high-value use cases in healthcare and finance. Real-time translation became broadly available for contact center text channels in the late 2010s and more deeply integrated into agent desktops by the early 2020s.
Zendesk is not the first company to bring real-time AI voice translation to contact centers. By 2024, point solution providers Krisp and Sanas were working with CCaaS providers, BPOs and enterprises, typically through a partnership and overlay model. In the most common deployment, an agent runs a separate desktop application that captures and processes audio through a virtual microphone.
Sanas also offers a Chrome extension that provides in-app translation for browser-based CCaaS users. It works only within Chrome and cannot intercept audio from a standalone desktop softphone.Market.us Scoopestimates that only about 20% of contact center agent desktops worldwide operate using Google Chrome thin-client technology, meaning that an in-app translation experience is not broadly available out of the box.
Both Krisp and Sanas offer developer SDKs for embedding their technology into a native calling interface. That generally requires custom work by an enterprise or BPO IT team, or professional services from the vendor. For the most part, companies using Krisp or Sanas must establish a direct commercial relationship with the translation provider, even when installing the product through a CCaaS marketplace.
Zendesk makes the case for native integration
What is different about Zendesk’s announcement is that it productizes the embedded translation experience as a native, Zendesk-branded capability seen below and, in this demo, without requiring a separate application, vendor agreement or a custom integration project.
“The goal is to enable our agents and our customers to speak in the language they know, no matter what language the other person is speaking and hear the language they know as well,” Shoemaker explained during our pre-briefing.
One major advantage of real-time translation is a reduction in expensive multilingual staffing. Companies could route interactions according to the issue and agent expertise instead of language, while supporting languages with too little demand to justify a dedicated queue.
“They install on the desktop as an app and basically it’s a virtual microphone,” said Nuri Gocay, director of Contact Center Platform Architecture, Zendesk. “The difference is that we’ve done this in our product as one piece. We’ve got all the context.”
That context includes contact attributes and call events. If a Mandarin-speaking customer first interacts with a Zendesk voice AI agent, for example, an English-speaking live agent can receive an English summary. The language attribute can automatically turn on translation for both sides.
“Admins don’t need to deal with configuring those additional tools,” Shoemaker said. “Agents have the context they need for the customer and why they came in.” Translation becomes part of the routing, context, call controls and recording workflow. That means smoother handoffs for agents and fewer integration and governance points for administrators.
Zendesk appears to be the first mainstream CCaaS provider to bring this capability directly into its agent desktop. Some BPOs and enterprises have created similarly embedded experiences, but they have generally done so by licensing an SDK and building the integration themselves.
