Yes — you can run a user-research interview with someone who doesn't share your language without booking an interpreter or forcing them into yours. A translation bot joins the call, the participant speaks in whatever language they think in, you read live captions in yours, and afterward you code and synthesize from one translated transcript. The reason to do it this way isn't convenience — it's signal. The whole point of a research interview is the nuance: the hesitation, the exact word someone reaches for, the offhand objection. Make a participant work in their second language and you sand all of that off before it reaches you. Here's how to run the session so it captures what the person actually thinks, not a flattened translation of it.
A second-language interview quietly destroys your signal
When you interview a user in a language they're not fluent in, what you lose isn't obvious — which is what makes it dangerous. They simplify. They give you the answer they can say instead of the answer they mean. They hesitate and round the specific down to the generic, because reaching for the precise word costs more effort than it's worth mid-sentence. You walk away with a transcript full of "it's fine," "it's easy to use," "no problems" — and you read that as a calm signal when it's actually a language ceiling. Research lives on the texture: the story about the workaround, the half-formed complaint, the word they keep returning to. A participant working in their second language can't give you that even when they want to, and you can't recover what was never said. Letting them speak the language they think in is the difference between a usable interview and a polite one.
Let the participant speak the language they think in
The fix is simple to state and easy to skip: let the participant talk in their own language. People narrate their real reasoning in their first language — the digressions, the analogies, the emotional weight on certain words — and that's exactly the material a researcher is there to collect. The objection is usually "but I don't speak their language," and that's what the bot is for. The participant answers naturally, and you read the same speech live in your language, in about two seconds. You're not waiting on an interpreter to relay a cleaned-up version, and you're not asking the participant to perform in a language that makes them sound less thoughtful than they are. Sageio translates into 20+ languages, so this isn't only for the major markets — it's whatever language your participant is most fluent in, read by you in whatever language you work in.
Read live, so you can follow the thread
A good interview is steered in the moment. You hear something and you follow it: "you said that was frustrating — tell me more." That probing is the skill, and it only works if you understand the answer as it lands, not five minutes later. Live captions keep you in the conversation: you read each answer as the participant speaks, in your language, fast enough to ask the obvious follow-up while it's still warm. That matters more in research than in most meeting types, because the value isn't in the first answer — it's in the third or fourth one you only reach by following the thread. Reading live means you can recognize the moment worth digging into and dig into it, instead of nodding along to a language you half-follow and realizing afterward you let the interesting part go by.
Code and synthesize from a faithful record
The live captions get you through the session; the transcript is what you build findings on. Within about five minutes of the call ending, a searchable translated transcript and an AI summary are ready — every answer, in writing, in your language. You can quote a participant accurately instead of paraphrasing from memory. You can tag and code answers — affinity-mapping, theme-counting, whatever your method — against the actual words rather than your impression of them. When you pull a verbatim into a readout, it traces to a record, not to "I think she said something like." And because the team works from the same translated transcript, two researchers coding the same interview are coding the same answers, which makes synthesis more consistent and findings more defensible when a PM pushes back. A faithful record is the whole game in research — the difference between evidence and anecdote.
How to do it with Sageio
- Add
bot@sageio.netto your Google Meet calendar invite. It joins on its own — no extension, nothing for the participant to install. - The participant and the researcher each pick a caption language. The participant speaks in the language they think in; you read yours — both from the same speech, at the same time. (Sageio translates into 20+ languages.)
- Run the interview normally. Translated captions appear in about two seconds, so you can follow the thread and probe in the moment.
- Afterward, a searchable transcript and an AI summary arrive within about five minutes, shared at the host's discretion — you code and synthesize from one record.
(Today this runs on Google Meet; Zoom and Microsoft Teams support is coming soon.)
Is it private?
For anything that joins your meetings: Sageio doesn't use your meeting content to train AI models, and its AI vendors are contractually restricted from doing the same. Audio is processed in memory and discarded — only the text transcript and summary are kept, encrypted, in the region you choose (US, EU, or APAC). The bot also joins as a visible participant, so tell the participant it's there and that it's translating — for research that's just good consent practice, and most people are glad to speak their own language.
Frequently asked questions
Does the participant have to install anything?
No. You add bot@sageio.net to the calendar invite from your side, and it joins the Google Meet on its own. The participant takes the call as they normally would — no extension, no account, no setup on their end.
What languages does it cover? Sageio translates into 20+ languages, so the participant can speak in the language they're most fluent in and you can read captions in yours.
Can I quote participants from the transcript? Yes — that's much of the point. The translated transcript gives you the answers in writing, so you can pull verbatim quotes and tag findings against the actual words instead of paraphrasing from memory. As with any research, get consent for how you'll use and share quotes.
How fast are the captions and the transcript? Captions appear in about two seconds, fast enough to follow the thread and probe live, with a searchable transcript and AI summary within about five minutes after the call ends.
What does it cost to try? Every plan starts with a free 60-minute trial, no credit card required. After that, Professional is $49/month and Teams is $99 per seat/month (annual billing includes 2 months free); Enterprise is custom-priced.
If your users don't all speak your language, you don't have to choose between skipping their input and interviewing them through a second-language filter that erases the signal. Let them speak the language they think in, read it live, and code your findings from a faithful translated transcript. For a related case — interviewing on the hiring side — see interviewing a candidate who speaks a different language, and for the broader picture, real-time translation for remote teams. Add the bot to your next interview and see how much more of the person comes through.