Answering customers in several languages

Where the time actually goes

In a single-language queue the cost of a message is the reply. In a mixed queue it is everything before the reply: opening the message, realising it is in a language you read slowly, copying it into a translator, reading the translation, going back, and only then deciding whether this is today’s problem or next week’s.

Multiply that by forty messages and the day is gone to context switching rather than to work. The fix is not better translation; it is knowing what is in the queue without opening every message in a second window.

What Mailbind does about it

The split between those first two points is deliberate: you want to understand in your language and answer in theirs. Nothing is sent automatically, and the draft always opens in the composer — a machine-written reply in a language you do not read is exactly the thing that should be read by a person first.

• A two-line summary at the top of every message, in the language you read the interface in, plus the next step if there is one.
• Suggested replies written in the language of the sender — French for a French customer, Arabic for an Arabic one — as drafts you edit and send.
• A rewrite tool that translates your own draft before you send it, and makes it shorter, friendlier or more formal.
• Translation of a received message, subject included, without leaving the reader.
• Needs-reply triage, which sorts the queue by whether a human is still owed an answer rather than by language.

An interface that mirrors, not a translated one

If part of your team reads Arabic, the interface itself matters. Mailbind in Arabic is mirrored throughout: the list moves to the right, the reader to the left, arrows and swipes reverse, and the layout is written in logical properties so direction is one attribute on the document rather than a second stylesheet.

Mixed text is handled rather than hoped for. Names, subjects, addresses and message bodies are each isolated in their own direction, so an Arabic subject containing an English product name reads the way its author typed it. The interface language belongs to the person, not the workspace, so three colleagues can read the same shared inbox in three languages.

What to tell your customers, and what not to

Two honest limits. First, a suggested reply is a draft produced by a model and it can be wrong, in any language; a reply in a language you do not read should be checked by someone who does before it becomes a habit. Second, the AI sees only the message you have open, and only while it is producing the result — which is also the answer to the question a careful customer will ask about where their mail goes.

AI can be switched off for the whole account or for one mailbox, which is the right setting for a client whose mail must not leave their provider’s country. The sub-processor list is published in the privacy policy, naming who processes what and where.

Common questions

Keep reading

• All use cases
• Run a shared support inbox
• Arabic and RTL email client
• AI email assistant
• For ecommerce support
• Security at Mailbind