Voice AI for non-English speakers, what's actually working in production
Curious what use cases people here have seen succeed with multilingual voice agents. From what I've been reading, healthcare appointment booking in Spanish is basically a solved problem at this, point, and Arabic is close behind, still some accent inconsistencies depending on dialect, but viable enough for production. Debt and EMI reminders in regional dialects also seem to be doing really well, especially across South Asian languages like Hindi, Tamil, Telugu, and Bengali. The hyperlocal support has gotten surprisingly deep. The auto language detection stuff is what I keep coming back to. Detecting from the first couple words and switching mid-call is pretty standard now, and the better systems are handling code-switching too, like someone bouncing between Hindi and English in the same sentence, which is just how a lot of people actually talk. That feels like a genuinely hard problem that's quietly gotten a lot better. I work with a lot of non-native English speakers and the trust angle is something I think about constantly. When someone is navigating healthcare or a loan reminder in their second language, the cognitive load is already high. A voice that sounds native to them, even if it's clearly an AI, probably changes the whole dynamic in ways that are hard to measure but really matter. There's also something to be said for culturally adapted prompts, not just the language but the framing and tone. Curious if anyone here has actually built or deployed something in this space. What was the hardest part to get right? Accent handling, latency, something else entirely?