Voice Notes and Languages You Do Not Speak
A guest sends a voice note in Georgian. Nobody on your team speaks Georgian. It arrives at 9pm.
Mushon Nachmani
vGuest
A guest sends a voice note in Georgian. Nobody on your team speaks Georgian. It arrives at 9pm.
What happens next tells you more about your guest communication than any dashboard.
The two assumptions that quietly limit most deployments
Most messaging setups are built on two assumptions that nobody states out loud.
That guests type. They often do not. Voice notes are the default input for a large group of travellers - older guests, anyone whose language is awkward on a phone keyboard, anyone standing in a street with luggage in one hand. Treating voice as an edge case pushes exactly those guests back to the phone, which is the channel the whole deployment was supposed to relieve.
That guests write in a language you handle. They write in whatever they think in. That set is determined by your actual guest mix, not by which languages your website offers, and it reliably includes several nobody on the team speaks.
Where those assumptions hold, messaging works beautifully. Where they break, the guest hits a wall - and it is invisible to you, because a guest who gives up does not file a complaint.
What good handling actually looks like
The functional requirement is unremarkable to describe and worth being precise about:
A guest sends a voice message in any language. The system transcribes it, understands the request, and replies in the same language the guest used - instantly, and in the property's tone rather than as flat machine output.
Three parts of that deserve attention.
Same language, without being asked. The guest should not have to select a language or request translation. They wrote in Polish; they get Polish.
Voice in does not have to mean voice out. A text reply to a voice note is usually right - it is skimmable, re-readable and easy to screenshot. What matters is that the voice note was understood, not that the format is mirrored.
Tone survives translation. This is the part that separates a useful system from a serviceable one. A translation that is accurate but reads as generic loses what makes the property distinctive. A guest writing in their own language should get the same warmth as one writing in yours - that is the whole point.
Why this matters more than it appears
The obvious benefit is coverage: you can answer guests you could not previously answer.
The real benefit is what it does to the shape of your inbound volume.
A property that cannot handle a language handles it anyway, badly and expensively. The message sits until someone finds a colleague who speaks it, or the guest is asked to write in English, or they give up and telephone - where the same language problem now occupies a staff member in real time.
Multilingual handling does not add a capability so much as remove a category of interaction that was consuming disproportionate effort. The guests who were hardest to serve become the easiest, because their questions are usually the ordinary ones - what time is check-in, is there parking, can we arrive late.
Where to keep a human firmly in the loop
Two cautions, because automated multilingual handling makes it tempting to route everything through it.
Complaints and sensitive situations. Tone matters most exactly where translation is least reliable. A frustrated guest in a language nobody on the team reads should escalate with the original message preserved, not just the translation, so whoever picks it up can see what was actually said.
Anything with legal or financial consequence. Cancellation terms, damage charges, medical or accessibility requirements. Translation is good; it is not a substitute for a human confirming that both parties understood the same thing when money or safety is involved.
The test worth running
Before assuming this works, send your own system a voice note in a language nobody on your team speaks. Then read the reply, and put it in front of someone who does speak it.
You are checking three things: whether it understood the request, whether the answer is correct, and whether it sounds like your property or like a machine. The first two usually pass. The third is where properties discover that their multilingual coverage is technically complete and tonally absent.
That gap is fixable, and it is worth fixing - because for a guest who cannot write in your language, the message they receive is the entire impression they have of your service before they arrive.
Common questions
Can an AI assistant handle voice messages from guests?
Yes. A guest can send a voice note on WhatsApp in their own language and receive a reply in that language, translated while preserving the property's tone. This matters because voice is the default input for a large group of guests - older travellers, people typing in a non-native script, and anyone with their hands full or standing in the street.
How many languages does a hotel realistically need to support?
More than it thinks, and usually not the ones on the website. The languages that arrive are determined by your actual guest mix rather than your marketing plan, and they include the ones no member of staff speaks. That is precisely where automated multilingual handling earns its place, because the alternative is a guest waiting for someone who never becomes available.
Does replying in a guest's language damage the property's voice?
It should not. The requirement is that the assistant translates while preserving tone and style rather than producing flat machine output. A reply that is correct but reads as generic is a missed opportunity - the point is that a guest writing in their own language gets the same warmth as one writing in yours.
Why do guests send voice notes instead of typing?
Because it is faster and easier, especially in a script that is awkward on a phone keyboard, and especially for older guests. Treating voice as an edge case pushes those guests back to the phone, which is the channel the messaging deployment was meant to relieve.
Keep reading
All articles
Guest Experience
The AI Protocol Shift: Why Your Hotel Messaging Must Become Machine-Readable
8 months ago · 8 min read

Guest Experience
Not all “Hotel AI” is created equal
9 months ago · 7 min read

Guest Experience
Beyond Check-In: How AI Guest Communication Transforms Hotels, Resorts, and Boutiques
9 months ago · 5 min read
190+ properties · 6 countries
See it answering your guests
Bring your own property. We will show you what the assistant does with your real questions, in your languages, on WhatsApp.