One knowledge interview runs in the language you write it in plus up to five more, all behind a single link. Your colleague picks theirs on the first screen and answers entirely in it — and the results still come back to you in one language.
The people who hold the knowledge worth capturing are often not the people the software was bought for. The credit controller in Madrid, the service-desk lead in Kraków, the claims analyst in Milan — asked in a second language, they give the version of the process they can say confidently, which is the tidy one. The exception, the hesitation and the reason behind a judgment call are the first things to go.
Updated · Nuggetz
Three settings, and that is the whole model
You are not managing translations. You are choosing what the interview is written in, who may answer it in something else, and what you want to read at the end.
Interview in
The language you write the guide in. You choose it when you create the interview, and the assistant drafts every question in it.
Participant languages
Up to five further languages a colleague may pick instead of the one the guide is written in. Add them on the draft; launching is what translates them.
Results in
The one language the summaries, nuggets and handbook are written in, whichever language each colleague answered in.
What the colleague actually experiences
A language choice on the first screen, then nothing else that reminds them they are using a translation.
The questions are in their language. So are the AI's follow-ups — asked in the moment, about what they just said, not pulled from a translated list. The read-aloud voice speaks their language, and when they answer by speaking, the transcription is told which language to expect rather than guessing. Arabic and Hebrew lay out right to left. Every language is translated in full and checked before the interview goes live, so the wording holds steady from the first question to the last instead of drifting as it goes.
Answers in six languages, one handbook
This is the part that matters more than the translation. Knowledge captured in six languages is six piles of notes until something reconciles it. Nuggetz writes every summary, nugget and handbook in the one language you chose, so an answer given in French and an answer given in Italian describe the same process in the same words — which is what makes it possible to say that two colleagues agree, or that they do not.
Quotes are the exception, and deliberately so. The colleague's own words stay exactly as they said them underneath the claim they support, because a translated quote is no longer evidence. Every nugget state — confirmed, single source, conflict or gap — is decided across all the languages at once, which means a conflict between a Polish answer and a German one is a finding you now get instead of a coincidence nobody was positioned to notice. The chain is the same one described on how it works.
Twenty-eight languages
العربية
Arabic
中文
Chinese
Čeština
Czech
Dansk
Danish
Nederlands
Dutch
English
English
Suomi
Finnish
Français
French
Deutsch
German
Ελληνικά
Greek
עברית
Hebrew
हिन्दी
Hindi
Bahasa Indonesia
Indonesian
Italiano
Italian
日本語
Japanese
한국어
Korean
Latviešu
Latvian
Norsk
Norwegian
Polski
Polish
Português
Portuguese
Română
Romanian
Русский
Russian
Español
Spanish
Svenska
Swedish
ไทย
Thai
Türkçe
Turkish
Українська
Ukrainian
Tiếng Việt
Vietnamese
A language is on this list only once the models that write it, read it aloud and transcribe it have each been checked against it. The list grows that way rather than by adding whatever a provider claims to accept.
What it does not do
Finding one of these out after the demo is a worse outcome than reading it here.
Nuggetz itself is in English. What gets translated is the interview your colleague sees — the builder, the dashboard and this site are not.
A colleague picks a language on the first screen and keeps it to the end. There is no switching mid-interview, and Nuggetz never guesses their language from what they say.
Five participant languages per interview on top of the one it is written in, chosen from the twenty-eight below. Regional variants — Brazilian Portuguese, Swiss German, traditional Chinese — are not offered separately, because no provider has confirmed them as distinct.
Translations are regenerated from the guide rather than edited by hand, and every language setting is fixed once the interview launches.
Edit a live guide and the affected translations are marked out of date until you refresh them. Nobody is ever served a half-translated interview: a colleague part-way through finishes the version they started.
Frequently asked questions
How many languages can one interview run in?
Six: the one you write the guide in, plus up to five more chosen from a list of twenty-eight. They all sit behind the same link — you do not build or send a second interview for a second language.
How does a colleague choose their language?
Buttons on the first screen, with their browser's language already selected. Everything after that is in the language they pick: the questions, the AI's follow-ups, the voice that reads them aloud and the transcription of what they say back. Adding ?lang=fr to the link preselects one, which is useful when you know who you are sending it to.
What language do the results come back in?
One language that you choose, whichever language each colleague answered in. A French answer and an Italian answer produce nuggets in the same language, which is the only way knowledge from both can be compared, corroborated and written into one handbook. Quotes stay in the words the colleague actually used.
Is the translation done live, while someone is answering?
No. Every language is translated in full when you launch, and checked before the interview goes live — so wording and terminology stay consistent rather than drifting question by question, and nobody waits mid-answer.
Can we interview people in Arabic or Hebrew?
Yes. Both are supported and the interview lays out right to left for them.
Is the Nuggetz app itself available in other languages?
No. The product you sign in to is English. Multilingual support is about the colleague answering the interview, who often is not the person who bought the software.
Ask the person who knows, not the person who can explain it in English
Pick one process your team depends on, and one colleague who has never been asked about it in their own language.
