Rumors of the demise of voice in the contact center appear to have been greatly exaggerated.
In fairness, those rumors were supported by fairly solid logic: new digital channels would become more and more popular, making the phone a costly legacy channel.
But that isn’t what’s happened.
In fact, Zendesk says voice still accounts for around 40% of contact center contacts among its customers, with some even reporting a 24% increase in voice volume on average.
“Voice is more important than ever,” said Hannah Deveney, Senior Director of Product Management at Zendesk.
So why hasn’t digital managed to kill off voice?
In a nutshell, the two channels are working together, rather than against each other.
AI agents are beginning to take the simpler calls, leaving human agents to deal with the disputes, fraud claims, billing problems, and sensitive conversations that customers do not want to resolve through a string of messages, as Deveney explains:
“When a call is escalated to a human, it’s usually a more complex or sensitive issue. Humans want to speak to humans to have that empathy or care when it’s a really complicated issue and address it in real time.”
Voice Translation Is About More Than Language
This need for phone channels to be able to provide a more personal customer experience was part of the thinking behind Zendesk’s latest Real-Time Voice Translation feature.
Built directly into Zendesk contact center, the tool translates live calls in both directions.
The capability is designed to allow customers and agents to speak naturally in their preferred languages.
Agents can turn translation on or off during a call, while administrators can decide where it is available and whether to retain original audio, translated audio, or both.
On the surface, that sounds like a useful way to cut interpreter costs and reduce pressure on multilingual hiring. Yet the more interesting opportunity is around routing.
Traditionally, a multilingual customer may be directed to a language-specific queue, regardless of whether that agent is the best person to handle their issue.
A French-speaking customer with a complex fraud claim, for example, may be passed to a French speaker who has broad service knowledge rather than a specialist trained for fraud cases.
That creates a difficult trade-off between language and expertise.
“You have your specialists who might handle billing inquiries and then others who handle sales. But then language gets its own queue,” Deveney said.
“You’re kind of creating these isolated silos when you want to have your agents who are proficient and excellent at solving a specific need to be able to speak the language as well.”
With real-time translation, instead of routing primarily by language, contact centers can route based on the customer’s issue and the agent’s specialist knowledge, then use AI translation to keep the conversation moving.
For customers, the payoff should be fewer transfers, less repetition, and a better chance of reaching someone who can resolve the problem first time.
Keeping Context When the Stakes Are Higher
Another advantage of real-time translation is the ability to keep the full customer story intact, even where the call crosses languages, as Deveney explained:
“What we’re trying to do here is have that customer context kept all the way through, even if it’s a multilingual call.”
We will all have experienced the irritation of being transferred between agents/departments and having to repeat information. Imagine how much more frustrating that would be when you add a language barrier into the mix.
Zendesk’s vision is clearly appealing. Contact centers have spent years trying to bring customer data, interaction history, and channel context into one workspace; translation could become another piece of that effort, rather than a separate service layer that breaks the journey.
However, making an interaction understandable is not the same as making it safe.
Translation Accuracy Is Not the Same as Trust
The biggest question is what happens when translation is used in high-stakes conversations.
An agent dealing with a fraud claim, health issue, formal complaint, or financial dispute may need to act on language they cannot independently understand. A technically accurate translation may still miss a cultural nuance, a phrase with legal significance, or a customer’s emotional state.
Zendesk is not suggesting customers should deploy the technology everywhere on day one.
“Don’t necessarily put your new multilingual human agent who can now speak any language on your most risky, most sensitive, high-stakes customer flows first,” Deveney said.
“Start small and put really good guardrails around what you’re doing.”
That is sensible advice for contact center leaders.
Start with lower-risk use cases; assess whether translation is improving resolution quality, customer effort, repeat-contact rates, and agent confidence; keep transcripts and recordings where appropriate, establish clear escalation paths; and make sure quality assurance teams can review the outcome.
Zendesk’s ability to retain original and translated audio could help with that audit trail. Still, organizations will need to test the technology against their own languages, dialects, compliance obligations, and customer scenarios.
Empathy Is More Than a Friendly Tone
There is another risk. Contact centers can become too excited by the idea of removing a language barrier and overlook what customers are actually asking for.
They want to be understood. They want the right person. They want a resolution without being pushed into a slower channel because the business cannot support their preferred language on the phone.
“You’re meeting them in the channel that they want to contact you on, you’re not deflecting them to another channel,” Deveney said. “It’s in real time.”
That is where voice translation could have its biggest customer experience impact.
AI cannot automatically give an agent cultural fluency, judgment, or empathy. It can, however, remove some of the friction that prevents an agent from using those skills.
If the customer’s context is preserved, the agent has specialist knowledge, and the call happens at the moment the customer needs help, the experience has a better chance of feeling personal.
For contact center leaders, the goal should not be to replace multilingual expertise with a translation feature. It should be to make sure language no longer decides whether a customer gets a fast, informed, and human resolution.
