Five interviews,one page of insights
Discovery interviews produce the most valuable text a product team owns, and the most scattered: typed notes per call, screenshots of prototype reactions, survey verbatims you can't select. Khint holds the whole study in one local work session, digests each interview with a saved AI action, synthesizes themes across participants, and publishes the findings to Confluence, before the calls fade from memory.
The insight decays faster
than the synthesis gets scheduled
By Friday you remember the last participant vividly and the first one barely at all. The raw material is all there (notes, screenshots, a colleague's observations), but it's in five places, half of it is pixels, and the synthesis workshop keeps slipping. Khint's three pillars close that gap together: Memory holds the study in one session, Capture reads the evidence that lives on screens, and Agents do the per-interview digesting and the cross-interview synthesis, all behind one palette shortcut. This page is about planned discovery research: interviews you ran, synthesized into themes. For sweeping unsolicited inbound signal into backlog items, see customer feedback to backlog instead.
The same study, two ways
What changes when the research has a home.
- Five interviews, five note files, three screenshot folders
- Retype the survey verbatims you couldn't select
- A synthesis workshop that fits nobody's calendar
- Findings shipped two weeks after the last call
- Every note and capture logs into one local session
- Drag a box over the evidence that lives in pixels
- One Action names the themes, with who said what
- The summary lands in Confluence the same afternoon
From five calls to one research summary
A discovery week, step by step.
Open one session for the whole study
Before the first call, start a Memory session named after the study, say Checkout discovery · June. Every note you process and every screenshot you capture from here on is logged into it, and its compacted summary follows you into the AI actions you run later. Five interviews stop being five files in five places and become one thread, on your machine.
After each call, digest the raw notes with one Action
Save an Action once: from these interview notes, list pains, verbatim quotes, feature requests, and anything surprising (type that as a one-line description and the Generate button drafts the prompt for you). When a call ends, select your raw notes wherever you typed them, hit the shortcut, and file the clean digest. Ten minutes of tidying becomes one keystroke while the call is still fresh.
Capture the evidence that isn't text
The participant reacted to a prototype screen; the survey tool shows verbatims in a panel you can't select; a colleague sent their observations as a photo of sticky notes. Pick Extract text in the palette and drag a box around it, and the words join your study as plain text. To question an image instead of copying it, Capture & ask opens a chat about the region you selected. Ask AI about your screen →
Run the cross-interview synthesis
This is the step that usually dies in a sticky-note wall. Save a second Action: across these interviews, name the recurring themes, note which participants support each one, and flag where they contradict each other. Run it with session context enabled, and the model sees the whole study (the digests, the captures), not just the note you selected. Recurring pains surface with receipts: who said it, in which interview, in their own words.
Publish the findings where the team reads them
With Atlassian connected on the integrations page, run Create page from the palette: pick the space, pick the parent page, and Khint drafts and files the research summary in Confluence. If your team keeps one living research page per product area, Append to pageadds this week's findings to it instead: safely versioned, so a colleague editing the same page at the same moment can never be overwritten. More on writing to Confluence →
What participants told you stays on your machine
Interview notes are personal by nature: names, workplaces, frustrations people shared because you promised discretion. In Khint, the session that holds your study lives in a local database on your machine and is never synced to Khint's servers. When you run a digest or synthesis Action, only the text you selected plus the session's compacted summary is sent for processing, and any Action can opt into PII redaction, which swaps emails and phone numbers for placeholders before the text leaves your device. The only thing that reaches your team's tools is the summary you deliberately publish. How PII redaction works →
Common questions
Does Khint record or transcribe my interview calls?
No, and that's deliberate. Khint is not a call recorder: it never joins your meetings and never touches audio. It works on what you already produce during a study: the notes you type, the prototype screens and survey panels you screenshot, the observations a colleague sends you. You stay in control of what enters the study; Khint's job is to hold it in one session and do the synthesis work on top.
Half of my evidence is screenshots: prototype reactions, survey verbatims. How does that become text?
Through Capture. Open the palette, pick Extract text, and drag a box around the verbatim panel or the sticky note in the screenshot; the words land on your clipboard as plain text and the capture logs into your active session. If you want to interrogate an image rather than copy it (say, a screenshot of a participant's messy workflow), pick Capture & ask instead and ask questions about the region you selected in a chat.
How does the AI see all five interviews when I run the synthesis, not just the note I selected?
That's what a Memory session does. Every note you process, every capture, every pull during the study is logged into the session, and Khint keeps a compacted running summary of it. When you run an Action with session context enabled, that summary rides along with your selected text, so a synthesis prompt like 'name the themes that recur across participants and flag contradictions' actually has the week of interviews to work with, not just the paragraph under your cursor.
Interview notes are personal data. Where does all of this live?
The session that holds your study lives in a local database on your machine; sessions are never synced to Khint's servers. When you run an Action, only the text you selected plus the session's compacted summary is sent for processing. For extra caution, any Action can opt into PII redaction, which replaces emails, phone numbers and similar identifiers with placeholders before the text leaves your machine. What reaches Confluence is only the summary you deliberately publish.
Ship the findings while the calls are fresh
Free with 10 AI actions and 5 captures per day. No credit card. One session per study, and the synthesis stops waiting for a workshop.