An AI interviewer runs the conversation for you. It asks a question, listens, and follows up based on the answer, so you can run qualitative interviews at a scale you couldn't personally staff.
When to use one
When you need to understand reasoning from more people than you could interview yourself. Thirty conversations, no scheduling, running around the clock across time zones.
Use a moderated interview instead when the subject is sensitive, the participants are very senior, or you'll want to improvise well beyond your guide.
Writing the guide
This is the part that decides whether the study works.
Write objectives, not a script. "Understand how they chose their current vendor" works better than a fixed list of questions, it lets the interviewer follow what the person actually says.
Three to five topics is plenty for a thirty-minute conversation. More than that and each gets rushed.
Each topic takes questions and probes. The questions are what it asks; the probes are the angles you want chased when an answer is interesting. Filling in the probes is what stops follow-ups being generic.
Open questions only. Anything answerable with yes or no wastes a turn.
Say what a good answer contains. If you need concrete examples rather than opinions, put that in the guide.
The builder will draft a guide from your study description, sized to the length you've set.
Audio or video
Interviews run on audio by default. You can switch them to video, which records the participant on camera as well.
Audio is the better default for most studies, people speak more freely without camera pressure, and you'll usually get a higher completion rate. Choose video when seeing the person genuinely adds something, like reactions to a design.
How hard it pushes
Follow-up depth is the control that most changes what you get back:
Depth | Follow-ups per topic | Use when |
Light | 1 to 2 | You have a lot of ground to cover and want breadth |
Moderate (default) | 3 to 4 | Most studies |
Deep | 5 to 7 levels down | The why matters more than coverage (expect longer conversations) |
If interviews are coming back thin, this is usually the setting to change before you rewrite your guide.
Setting up the interviewer
Language: fifteen available, including Spanish, French, German, Portuguese, Chinese, Japanese, Hindi and Korean. Participants are interviewed in their own language, with native voices rather than an accent.
Voice: twenty to choose from in English, each with a different character. Preview them before you pick.
Style: Conversational for a relaxed feel, Structured to stay close to your guide, or Probing to push harder on every answer.
Pace and tone: adjustable if the default feels rushed or flat.
Opening line: you can write the first thing the interviewer says. Worth doing: a sentence explaining who you are and why you're asking sets a better tone than a generic greeting.
Background context: notes on the persona and tone to adopt. Participants never see these, but they shape how it behaves.
Interruptions: by default the interviewer acknowledges someone cutting in, the way a person would. You can switch it to listen silently instead, which is faster but feels less natural.
Preview the interviewer before launching. Hearing it is worth more than reading about it.
Recording
These sessions always record. It isn't optional and can't be turned off.
The recording and transcript are how the conversation reaches you, so recording is the study rather than an addition to it. Participants are told plainly before they begin and agree before anything starts.
What you get back
Audio, a full transcript, and analysis across every session, themes that recur, quotes worth using, and sentiment.
Getting more from it
Interview more people than feels necessary. The marginal cost of one more is low, and themes only become trustworthy once you've heard them repeatedly.
Read three or four transcripts in full before looking at the summary. It calibrates you on what people are actually saying, and you'll spot things worth chasing in the next study.