The scheduling tax
Count the steps between "we should talk to users" and the first transcript in a traditional study: write the screener, send invitations, chase replies, offer slots, handle time zones, confirm, remind, absorb the no-shows, run the sessions, transcribe, then code. Most of that is logistics, and logistics is why a "quick round of interviews" takes a month and why so many teams quietly stop doing them.
Asynchronous interviews: how they work
You write the guide once. An AI interviewer runs it, one participant at a time, whenever each participant opens the link. There is no slot to book because there is no human moderator whose time has to be matched. Ten people can be interviewed at 2 a.m. in three countries simultaneously, and all ten transcripts are waiting when you log in.
The conversation is still adaptive: the interviewer reads each answer and probes before moving on, within the follow-up limit you set. It is not a form.
What changes for participants
- They answer when it suits them, typically in 10 to 15 minutes, on a phone or laptop, with no download or account.
- They type by default. No camera, no microphone, no finding a quiet room. Voice is there if they prefer to talk.
- They answer in their own language; the interviewer adapts.
- They see who is asking and why before they start, and consent explicitly.
What changes for researchers
- Recruiting becomes distribution: one link in an email, a product banner, a community post, or a panel like Prolific.
- No note-taking and no transcription step. Transcripts are editable text from the start.
- Analysis runs across the whole set: themes with quotes, sentiment by topic, drop-off per question, exportable as CSV or Markdown.
- You get notified as each interview completes, so you can read the first ones while the rest are still coming in and fix a bad question early.
Keeping the depth
The fear with async is shallow answers. Three things protect against it: a clear goal for the interviewer so follow-ups are pointed; questions about specific past events rather than opinions; and a pilot where you take the interview yourself and feel where it drags. Ask for one story per question and you will get paragraphs, not sentences.
A one-week plan
- Monday: write six questions and a one-line goal. Create the project. Interview yourself and one colleague; fix anything awkward.
- Tuesday: send the link to 40 to 60 people with a short note and a small thank-you.
- Wednesday to Thursday: read transcripts as they land. Add a follow-up instruction if a theme is emerging that the guide does not probe.
- Friday: with 15 to 25 completed, read the themes and their quotes, export the Markdown report, and share it with the team.
- AI User Interviews is built for this cadence and is free to start: Free plan with 5 active projects and 100 participants a month; Pro is $25 a month.