Live & answering now: +1 289-778-4594 β€” hear the Botnira AI receptionist in action. Start free trial →
News

Zendesk Launches Real-Time Voice Translation for Contact Center Calls

Zendesk announced Real-Time Voice Translation for Zendesk Contact Center on September 10, 2026 -- an AI-powered capability that translates live phone calls in both directions in real time, so a customer can speak in the language that feels most natural to them while an agent responds in another, without a transfer, a hold, or a human interpreter on the line.

  • ⏱ 6 minutes
  • πŸ“… September 10, 2026

What Zendesk announced on September 10

Zendesk announced Real-Time Voice Translation for Zendesk Contact Center on September 10, 2026, in a newsroom release titled "Zendesk Expands Contact Center Voice Support Across Languages." The feature is a two-way, real-time translation layer built directly into Zendesk Contact Center rather than sold as a separate product or bolted on through a third-party interpreter service. During a live call, a customer can speak in their preferred language while the agent hears and responds in their own -- each side hears a translated stream in near real time, without the call being routed to a different queue or a human interpreter being patched in.

"Customers should not have to choose between speaking the language that feels most natural to them and getting the support they need," said Shashi Upadhyay, Zendesk's President of Product, Engineering and AI, in the announcement. "Real-Time Voice Translation helps businesses make that kind of voice support available to more customers, while agents stay focused on solving the problem." In a separate comment on the launch, Upadhyay added that the goal was to keep complex or urgent conversations with the same agent rather than routing them elsewhere: "When an issue is complex, sensitive, or urgent, people want to be understood, not routed somewhere else."

Source: Zendesk newsroom press release, September 10, 2026.

How it works, and what's supported at launch

Administrators control where the feature is active: it can be configured to translate a call automatically before it ever reaches an agent, or an agent can switch two-way translation on mid-conversation the moment a language barrier becomes apparent, without leaving the call. Admins also choose whether Zendesk retains the original audio, the translated audio, or both, which matters for quality review and compliance in regulated industries. At its early-access stage, the feature supports Chinese, English, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Portuguese, Russian, Spanish and Vietnamese -- 13 languages covering a large share of Zendesk's global customer base.

Zendesk is rolling the capability out through a closed Early Access Program starting in October 2026, limited initially to eligible customers running Zendesk Contact Center Native. Coverage from trade outlet CX Today notes some real constraints on the early-access build: translation quality and latency vary by language pair, accent and audio conditions, and the initial release excludes multi-party calls, live transfers, supervisor monitoring scenarios, and video calls with translation enabled. Those are the kinds of edge cases contact centers will be watching closely before deciding whether to expand the feature beyond a pilot.

Sources: CX Today; itBrief Asia.

Why Zendesk is targeting language, not just volume

The release leans on one specific figure to justify the investment: Zendesk's own research puts voice at roughly 40% of total contact center volume, even after several years of chat, email and self-service AI absorbing an increasing share of customer contacts. That statistic is the crux of Zendesk's pitch -- phone calls haven't gone away even as AI has taken over more of the customer journey elsewhere, and language mismatch is one of the few remaining reasons a voice interaction still gets escalated, transferred, or handed to a specialist. Historically, a contact center serving multilingual customers had three options: hire and staff separate language-specific agent pools, route calls to a smaller shared pool of bilingual agents (creating longer queues for less common languages), or bring in a human interpreter service mid-call, which adds cost and latency to every multilingual interaction.

Real-Time Voice Translation is Zendesk's attempt to collapse those three options into one: a single agent workforce that can, in principle, handle a call in any supported language without special routing. That reframes AI's role in the contact center less as a replacement for agents and more as a multiplier on which agents can handle which calls -- an agent's product knowledge or account access, not their language skills, becomes the deciding factor in whether they can take a given call. A CX Today weekly roundup grouping this release alongside contact center announcements from Genesys, RingCentral and Verint the same week noted that none of those competitors made an equivalent multilingual-voice claim in their own announcements, suggesting Zendesk is trying to stake out real-time call translation as a differentiator before rivals catch up.

Source: CX Today contact center weekly roundup, September 2026.

Why it matters

What this means for the voice AI industry

🌐

Language is becoming the next AI voice frontier

With transcription, summarization and routing largely commoditized in contact center AI, real-time bidirectional translation targets one of the few remaining hard problems in live voice support -- and could push rivals to ship comparable capabilities to avoid ceding multilingual accounts.

πŸ‘₯

AI as a staffing multiplier, not just automation

Rather than replacing agents, this positions AI as removing the constraint that ties a customer's language to a specific agent pool -- letting contact centers staff for skill and availability first, and worry about language second.

⚠️

Early-access limits show voice translation is still hard

The exclusion of multi-party calls, transfers, supervisor monitoring and video from the initial release is a reminder that real-time, low-latency call translation at production quality remains a genuinely difficult engineering problem, not a solved one.

Continue reading

Related reading

πŸ“°

Genesys Adds ElevenLabs and Deepgram Voice AI to Its Contact Center Agents

Another major contact center platform embedding more voice AI capability into its core product the same week, though focused on orchestration and voice quality rather than translation.

Read the article →
πŸ“°

Microsoft Launches MAI-Transcribe-2, Undercutting Meta and OpenAI on Price

A look at the underlying speech-to-text and multilingual model race that features like Zendesk's translation layer ultimately depend on.

Read the article →
🧭

AI Receptionist Research Hub

How enterprise contact center AI compares to the AI receptionist tools small businesses use for calls, booking and follow-up.

Browse the research →

See a small-business AI receptionist in action

Start your 5-day free trial -- no credit card required, live in about 3 minutes.