Will businesses get lost in AI translation?
Multilingual AI agents can now pick up the phone, hold a conversation, and switch languages on the fly. The promise is seductive: faster, cheaper support at global scale. The reality is murkier, and the stakes—customer trust, brand reputation, even legal exposure—are high. As AI crosses the language barrier, will it elevate service or erode the human connection customers value most?
Support roles in the AI blast zone
OpenAI’s Sam Altman has been blunt about where AI will bite first: customer support. In a September podcast, he predicted many call center roles would be replaced, arguing AI will handle phone and chat interactions “better.” He’s not alone. Salesforce CEO Marc Benioff revealed the company has reduced its support headcount from 9,000 to 5,000 and said its Agentforce bots should manage roughly half of all customer conversations by next year—while conceding large language models still have limits.
Translation: the tempting frontier
Language has long been a pain point in support. Misunderstandings lead to bad data capture, wrong outcomes, and frustrated customers who don’t get the help they need. AI vendors are racing to fix that. Google’s Gemini Enterprise bundle includes video translation that aims to go beyond literal word swaps, claiming to carry tone and expression across languages. Salesforce’s Agentforce Voice lets teams process calls with AI, tweak voice speed and tone, and fine-tune pronunciation of names or industry terms. And Sierra, from former Salesforce CEO and OpenAI chair Bret Taylor, pitches AI agents that can simply pick up the phone.
Where machines stumble: trust and context
Despite the momentum, many support leaders are pumping the brakes. John Campbell, senior vice president of client services at Map Communications, calls AI translation “super helpful” but warns of a fragile equilibrium. If a furious caller is misheard—or a phrase is mistranslated—the fallout can be permanent. “Imagine an extremely disgruntled customer on the phone with an AI agent, and something gets lost in translation. Not only is their trust going to continue to be damaged, but you could risk losing that customer forever,” he says.
Professional interpreters and translators have voiced similar concerns for years. Machine systems still struggle with ambiguity, idioms, formality levels, and cultural nuance—the subtleties humans use to navigate emotion, sarcasm, and context. A 2025 study by researchers at Shahjalal University of Science and Technology and the University of Oklahoma found that conventional machine translation often misses cultural and historical depth, with as much as 47% of contextual meaning lost. In high-stakes settings, these gaps aren’t academic.
The Guardian reported in 2023 that machine translation within the U.S. immigration system contributed to errors—such as “I” being rendered as “we”—and left some regional language variants, like Farsi Dari, unsupported. Transpose that risk to business and you get botched contracts, misleading support records, or noncompliant responses—issues that can spiral into reputational or legal crises.
The counterpoint: speed and scale
Still, the case for AI is strong. Cultural strategist and tech linguist Annalisa Nash Fernandez notes that while today’s models can misread context or overgeneralize, people make mistakes too. AI’s advantages—speed, cost, and 24/7 availability—are hard to ignore. Companies can stand up multilingual support quickly, expand access for diverse customer bases, and reduce wait times. Done right, AI translators can triage routine queries and free human agents to focus on complex, emotionally charged, or high-value issues.
What customers actually want
The catch: many customers don’t want AI front and center. A Kinsta survey of 1,011 U.S. consumers earlier this year found nearly half would cancel a service if forced to deal with AI-only support. Some 41.4% said AI has made service worse, and 41.3% would pay more to avoid AI agents altogether. “AI can mimic answers, but it can’t replicate the trust and growth that come from a truly human exchange,” says Roger Williams, partnerships and community manager at Kinsta.
Motivation matters. If companies deploy AI translation primarily to cut costs, customers will notice—and resent it. The goal, Williams argues, should be to enhance the experience, not replace the human relationships that make support meaningful.
The near-term playbook
For now, the winning strategy is balance. Put AI where it shines—quick answers, language bridging, form-filling, and guided troubleshooting—while maintaining a human-in-the-loop for edge cases, sensitive topics, and angry callers. Be transparent that AI is assisting. Offer easy escalation to a person. Invest in tuning: customize pronunciation, handle regional dialects, and test for idioms and cultural references. Above all, measure what matters: resolution accuracy, customer satisfaction, and the rate of AI-to-human handoffs, not just cost per contact.
AI may soon translate more than words—tone, pacing, and emotion are already on vendor roadmaps. But until systems reliably grasp context and cultural nuance, oversight isn’t optional. The question isn’t just whether AI can answer in every language—it’s whether customers will feel heard. Businesses that treat translation as a bridge to better human service, not a substitute for it, are far less likely to get lost along the way.
