AI analysis
Ten interviews make several hours of video. AI analysis lets you ask questions across all of them at once and get answers built from what people actually said, with every claim pointing at the moment in the recording it came from.
There are three tools, all inside a project: Ask AI, a chat over every transcript; Insights, a short list of findings drawn from across your sessions; and the Responses page, which lines up every respondent’s answer to one guide question. They read the recordings you made in LetsFindWhy and the ones you uploaded alike.
Say you have interviewed twelve developers about your CLI. Ask “What stops people from finishing setup?” and you get a summary with quotes from four sessions, each one a click away from the video. Then open the Responses page on the setup question to check that the other eight said nothing different.
Ask AI
- Open Ask AI in the project sidebar.
- Type a question about the research and click Send. While it works you see it search the transcripts and look up project data.
- Read the answer. Each claim ends in a small chip naming the speaker and the session. Under the answer, Quotes lists the passages it relied on word for word, and Sources lists each session it cited.
- Click a chip or a quote to watch that moment. The recording opens over the conversation and starts at the quoted line, so you can check the tone as well as the words.
Ask AI searches the transcripts of every session in the project, by meaning as well as by keyword, so “pricing” also finds “too expensive for a small team”. It can also look up the session list, who took part, their screener answers and the bookmarks you made, and it knows the project’s description and each process’s discussion guide. It cannot see survey answers or the rating and choice answers given in interviews; read those on the Responses page.
It answers in the language you ask in. Quotes stay in the language they were spoken, so an English question about Spanish interviews gets an English answer with Spanish quotes.
Every conversation is saved and named after its first question. Conversations are private: your teammates have their own and cannot see yours. Use the menu on a question to edit it or try again, which replaces the answers that came after it.
Insights
Insights are up to six findings drawn from across the project’s sessions, each one a short title and a sentence or two. Every insight has to rest on at least two passages from the transcripts, and one with no evidence behind it is thrown away.
- Open Ask AI. The insights sit under the greeting of a new conversation.
- Click Generate insights the first time. Later, a New data since last generated badge tells you there are new transcripts; click Refresh insights to rebuild the list.
- Click Dive deeper on an insight to ask about it in the chat, where the answer comes with its quotes.
Insights never refresh on their own after an interview, so you choose when the list changes. Everyone in the project sees the same list, and a refresh replaces it for all of them. Insights are currently written in English, whatever language the interviews were in.
The Responses page
Ask AI tells you what it found. The Responses page shows you everything, question by question: pick a question from your discussion guide and read every respondent’s answer to it, each with a clip to play. Rating and choice questions get a chart you can filter and split into groups, and Export CSV downloads those answers alongside each person’s own words.
It also lists who has no answer, and why: not asked, nobody responded, still transcribing. That is what a chat cannot give you, so use the two together. The full walkthrough is in Transcripts.
Rating answers
Every answer and insight has a thumbs up and a thumbs down. A thumbs down asks what was wrong (inaccurate, wrong citation, missed data or unhelpful) and takes an optional comment. Your rating is yours alone; click the thumb again to remove it.
Tips for researchers
- Ask the way you would ask a colleague who sat in on every session: “Which onboarding step did people complain about most, and how did teams on the free plan describe it?” gets a better answer than “summarise”.
- Treat an answer as a lead, not a finding. Play two or three of its quotes before you put it in a report; it takes a minute and you will hear what the transcript leaves out.
- A session can only be quoted once its transcript is ready. If you have just finished an interview and Ask AI says it cannot find it, wait for the transcript and ask again.
- Anyone who can view the project can use Ask AI, read-only viewers included. Asking questions does not use interview minutes.
- When your retention window deletes a recording, Ask AI can no longer quote it, and insights that relied on it are removed and rebuilt from what is left.