Zendesk will launch real-time voice translation for its Contact Center platform through a closed Early Access Program in October. Voice already accounts for 40% of contact center volume, but language barriers force customers into slower channels or specialized queues - a friction point that directly hits resolution times and satisfaction scores.
"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 President of Product, Engineering and AI.
How the translation tool works during live calls
Agents can toggle the feature on or off mid-call, translating conversations in real time without routing customers through language-specific queues. The tool removes the need to push urgent voice inquiries into email or SMS simply because no agent shares the caller's language. Administrators control where and how translation services appear, letting teams match the feature to their operational demands and language preferences.
This configurability matters for cost-conscious support teams. Rather than staffing for every language a customer might speak, a smaller team can handle a wider range of calls. The agent stays focused on solving the issue while the translation layer handles the language gap.
What the data says about voice support
Zendesk's own research puts voice at 40% of contact center volume. When language mismatches force those callers into asynchronous channels, wait times grow and first-contact resolution drops. For support leaders tracking CSAT and handle time, any tool that keeps conversations in the customer's preferred channel - and preferred language - removes a known bottleneck.
Smaller teams feel this pressure acutely. Hiring multilingual agents for every language combination is rarely feasible. The translation feature lets a lean team serve a linguistically diverse customer base without adding headcount or outsourcing calls to third-party services.
Rollout and practical considerations
The Early Access Program goes live in October for eligible Zendesk customers. As with any real-time AI translation tool, support managers will need to monitor output quality during live interactions. Training agents to work alongside machine translation - knowing when to confirm meaning, how to pace speech, and what to do if a translation misses the mark - will determine whether the feature improves outcomes or adds a new layer of complexity.
Integrating the tool into existing workflows also requires planning. The toggle-on-demand design helps, but teams still need clear protocols for when and how agents activate translation during a call. Done well, it becomes invisible to the customer. Done poorly, it creates pauses and confusion that erode trust.
Why this matters for customer support teams
Voice translation shifts the economics of multilingual support. Instead of staffing for language coverage, teams can staff for expertise and let translation handle the linguistic variety. For support leaders, the immediate metrics to watch are average handle time for multilingual calls, first-contact resolution rates across language groups, and CSAT scores from customers who previously had to switch channels because of language barriers. If those numbers move in the right direction, the business case writes itself - lower staffing costs, fewer escalations, and faster resolutions for a broader customer base.
Your membership also unlocks:
