A short-let host in Lisbon gets a WhatsApp message at eleven at night, in French, asking whether the flat is free the weekend after next and whether it is far from the airport. He does not read French. He copies it into a translator, works out roughly what is being asked, writes a reply in English, and sends it the following morning. No answer ever comes back.
He assumes the dates did not work. What actually happened is that a message written in French got an English reply eleven hours later, and the person reading it concluded, quite reasonably, that this was not a place set up for guests like her.
Who This Actually Hits
This is invisible if all your customers are local, and constant if they are not. The businesses it affects most are the ones where the customer base is geographically or linguistically mixed by default.
- Hotels, guesthouses and short-let hosts, where the entire market is by definition people from somewhere else.
- Clinics and dental practices that take patients from abroad, which is an increasingly ordinary thing for a practice to do.
- Businesses in border regions and tourist towns, where two or three languages in a week is a slow week.
- Anyone serving a neighbourhood where a second language is normal, which is most cities.
The pattern is the same in all of them. The enquiries arrive, they are harder to answer than the local ones, so they get answered later or worse, and the loss shows up as nothing at all, because a lead that never replies leaves no evidence of why.
What Happens to a Message You Cannot Read
There are only three realistic outcomes when a message arrives in a language nobody in the business speaks, and none of them are good.
It gets ignored, on the assumption it is probably spam or not serious. It gets answered in English, which sometimes works and often reads as a business that could not be bothered. Or it gets run through a translator and answered slowly, which is the most conscientious option and still adds hours to a response that needed to be fast to be worth anything.
The delay is the part that gets underestimated. A tourist comparing three hotels is not going to wait a day to find out whether one of them can accommodate them. They are picking from whoever answers tonight.
Translating Is Not the Same as Answering
Machine translation is genuinely good now, which makes it easy to assume the problem is solved. It is good at meaning. It is unreliable at register, and register is most of what makes a business message land properly.
A translator does not know that German has a formal and an informal you, and that using the wrong one with a stranger enquiring about a medical appointment is a real mistake rather than a stylistic one. It does not know that a French reply that is literally correct can read as curt to the point of rudeness. It does not know that in Italian a business writing to a new customer chooses its form of address deliberately. It will produce something grammatical and socially off, and the reader will not tell you, they will simply not book.
There is a second, quieter problem. Translating your reply after the fact means the conversation is only ever half in the customer's language. The booking confirmation, the follow-up, the reminder, the small clarification two days later, all of it reverts to English the moment anyone stops actively translating. A conversation that begins in someone's language and drifts out of it is worse than one that never pretended.
The French message about the flat arrives at 23:10. Within a minute there is a reply in French confirming those dates are free, giving the actual travel time from the airport and the price for two nights, and offering to hold the dates until the morning. The exchange that follows, including the confirmation, stays in French throughout. The host reads a short summary in English the next day and does nothing else.
Doing It Properly
The reason this is now practical is the same reason modern responders handle unexpected phrasing at all. A system that reads language rather than matching keywords does not need a separate build per language. It is the same responder, reading the same business information, replying in whatever it was written to in.
Four things need to be true for it to hold up.
- Detect from the message, do not ask. Making somebody choose a language from a menu before they can ask a question is friction, and it is unnecessary when the message itself already says which language it is in.
- Stay in that language for the whole conversation. Including the booking confirmation, the reminder and any follow-up. Consistency is the point, not the first reply.
- Keep the facts identical across languages. Prices, availability and policies come from one source. A responder that improvises a price in Italian because it was not given one has created a much larger problem than a slow reply.
- Hand over with a summary in your language. When a person needs to take the conversation, they should get what was asked and what was answered in the language they work in, without having to translate a thread themselves.
Common Questions
Can one responder handle several languages?
Yes. It reads the incoming message, works out the language, and replies in it from the same business information. Adding a language is not a rebuild, which is the main practical difference from older chatbots.
Is machine translation good enough on its own?
For working out what somebody wants, usually. For replying as a business, often not, because it preserves meaning but not register, and produces replies that are correct and socially wrong.
What if the conversation needs a human who does not speak the language?
The handover should carry a summary in the language your staff work in. The customer keeps being answered in theirs, your team reads it in yours.
How much of this do I actually need?
If more than a message or two a week arrives in another language, it is already costing bookings, and the fix is the same setup you would want anyway with the language handling switched on.
Where to Start
Scroll back through your inbox and count the messages that were not in your language. Then count how many of them turned into a booking. That second number is usually the one that decides this.
LeadOro builds multilingual response into the system itself, so enquiries get answered in the language they arrive in, at the same speed as everything else, with your real prices and availability behind every reply and a clean summary for you when a person needs to step in.
- Setup time: 7–14 days from the first call.
- Pricing: plans start at $500 setup + $150/month, month-to-month, no contract.
- Guarantee: 14-day satisfaction guarantee.
The enquiries in other languages are not harder customers. They are the ones nobody else in your area is answering properly either.
Answer every enquiry in the language it arrived in.
Book a free 15-minute discovery call. Tell us which languages turn up in your inbox and we will show you how the conversation would run in each of them.