AI agents handling routine queries has fundamentally reshaped voice's role in contact centers, contrary to predictions of its decline. Zendesk data shows voice still accounts for 40% of customer contacts, with some operations reporting 24% increases in call volume. The channel has not been displaced by digital alternatives; instead, the two work in tandem. AI now absorbs simple, transactional calls, leaving human agents to manage disputes, fraud claims, billing problems, and sensitive conversations where customers explicitly demand real-time human interaction. This division of labour has made voice strategically more valuable, not less—it has become the channel where complexity, risk, and emotional stakes concentrate. For teams already running Agentforce or similar agentic platforms, this shift means voice infrastructure and agent training must evolve to handle exclusively high-friction scenarios, fundamentally changing how you measure agent productivity and quality.
Zendesk's real-time voice translation feature addresses this new reality by removing language as a routing constraint. Rather than directing multilingual customers into language-specific queues regardless of issue type, teams can now route by expertise and use AI translation to bridge the gap. The operational benefit is clear: fewer transfers, reduced repetition, and higher first-contact resolution rates. More importantly, the feature preserves customer context across multilingual interactions and retains original audio for compliance and quality assurance. However, the critical tension lies in deployment scope. Translation accuracy is not equivalent to trust in high-stakes conversations—a technically correct translation can still miss cultural nuance, legal significance, or emotional tone. Zendesk explicitly advises starting with low-risk use cases and building guardrails before expanding to fraud, health, or financial disputes.
The deeper implication for CX leaders is that real-time translation succeeds only when it removes friction without replacing expertise. AI cannot manufacture cultural fluency or judgment; it can only eliminate the operational barriers that prevent agents from applying those skills. If your team treats translation as a cost-reduction tool to hire monolingually, you will fail. If you treat it as infrastructure that allows your best specialists to serve customers in any language without transfer, you unlock genuine competitive advantage. The question becomes whether your quality assurance and governance frameworks are mature enough to validate translation outcomes in your specific languages, dialects, and compliance contexts before deploying at scale.
AI Is Taking the Easy Calls. Can It Make Voice More Human? cxtoday.com
AI Is Taking the Easy Calls. Can It Make Voice More Human? CX Today