A company translates its website into four languages. It is done well, it costs a reasonable amount, and traffic from those markets increases immediately.
The enquiries arrive in four languages into an inbox staffed by people who read one. Each is translated roughly, answered in English, and sent two days later.
The translated site did its job. It brought people to the point of asking, which is precisely where the business could no longer serve them, and the investment produced frustration rather than revenue.
Who currently handles a non English enquiry
Whoever opens the inbox, with a translation tool, when they have time, which is a process nobody designed and everybody tolerates.
Who should handle it
Something that reads and replies in the customer's language immediately, hands to a person where judgement is needed, and gives that person a translated summary.
What translation actually buys
Discovery and credibility. Somebody finds the site, understands the offering and decides to make contact, which is genuinely valuable and is the first step rather than the whole journey.
What translation does not buy
The conversation, which is where price, suitability and terms are established, and where the business is actually won or lost.
What a customer concludes from an English reply
That the translated site was marketing rather than service, which damages trust more than never having translated anything.
What the response time does
Compounds the problem, since translation delay adds to the usual delay and the customer experiences both as indifference.
Where the formality problem sits
In languages with distinct registers, an inappropriate form reads as rudeness, which is the detail covered in answering in the language the visitor used.
Where a person is still required
Anything involving a commitment, a price negotiation or a sensitive matter, and the handover should carry a summary rather than a raw thread.
Where the internal speaker fits
Businesses frequently have somebody who speaks the language, and the design should route to them rather than leaving it to chance.
How to choose which languages
From traffic and enquiry data rather than assumption, which is the ground in deciding which languages to support first.
What to do about the sensitive content
Prices, terms, legal statements and safety information need human review in every language, and should never depend on automatic translation alone.
What consistency requires
The same answer in every language, maintained from one source, or the versions drift apart within months and nobody notices.
What the measurement should be
Not traffic from a market but enquiries answered and converted in that language. Businesses celebrate the first and never look at the second, which is why a translated site can appear successful while producing nothing.
What happens to the content over time
English pages get updated and the translations do not, so within a year the four language versions describe a business that has changed. One maintained source of answers avoids the whole problem.
Translate the answering, not only the pages
An assistant that reads and replies in the visitor's language, answers the routine questions accurately, maintains one set of answers across every language and hands to a person with a translated summary makes a translated website into a business that can actually serve those customers. The approach is set out in multilingual chatbots how ai answers in your customers language, and a sector example in answering patients in their own language. Start at try, see features, and the sector view sits under small business.