UAT feedback in,clean defect tickets out
A UAT round buries the business analyst in feedback that isn't a bug report: email threads, screenshots of error dialogs, “the screen looks wrong”. Khint turns the triage into a loop you run from one shortcut: capture the feedback, classify it with a saved AI action, file the real defects to Jira, and let a Memory session keep the whole round in context for the closing summary.
Most UAT feedback
isn't a bug report
Business testers are not QA engineers, and acceptance testing is not a test suite. Feedback arrives as a reply-all email, a cropped screenshot with no caption, a comment dropped in the hallway after the demo. Some of it is a genuine defect; a lot of it is a change request wearing a bug costume, or a training gap that no code change will fix. In a time-boxed round, the BA becomes a human parser: re-typing feedback into a tracker spreadsheet, deciding what each item really is, and rewriting the defects so a developer can act on them. That sorting work is exactly what a saved AI action does well, and because Khint runs it on selected text in whatever app the feedback landed in, the triage happens where the feedback already is.
One triage loop, from raw feedback to filed defect
Capture, classify, file, retest: all behind Cmd+Shift+K.
Open a Memory session for the round
Before the first tester logs in, start a Memory sessionnamed for the round (“UAT round 2: Billing”). From here on, every capture, every pulled email thread, and every action you run logs into it, and its compacted context rides along with your later actions. The round accumulates in one place instead of across your head, a spreadsheet, and six browser tabs.
Get every piece of feedback into text
UAT feedback never arrives in one shape. A tester replies in an email thread: pull it into the session with the palette's Gmail Pull thread row. Another sends a screenshot of an error dialog: run Extract text from Capture, draw a region, and the text is on your clipboard. A confusing screen? Capture & ask opens a chat about the image so you can work out what the tester actually hit.
Classify before you file
Most UAT feedback is not a defect. Save a triage Actiononce (“classify this feedback as defect, change request, or training gap, and say why”) and run it on each raw item from Cmd+Shift+K, wherever the text lives. The classified version pastes back in place. “The button looks wrong” becomes a decision, not a debate in the go/no-go meeting.
File the real defects without opening Jira
For items triaged as defects, the palette's Jira Create issuerow does the filing: pick one of your agents to draft the ticket from the triaged text (or use the text as-is), review the preview, confirm, and the new issue key comes back and logs into the session. Because the drafting agent can see the session's context, the ticket inherits what the round already knows.
Run the retest loop from the same shortcut
When a fix lands, the palette's Add comment and Change statusrows close the loop: pick the issue from your recent list, note the retest result, move the status. At the end of the round, select your notes, run a summary action over the session's accumulated context, and publish it as a Confluence page or send it by email: preview first, then confirm.
A Thursday of UAT, end to end
9am: the round opens, you start the “UAT round 2” session. 10am: three email replies from testers. Gmail Pull thread logs each into the session. 11am: a screenshot of a validation error. Extract text pulls the message out of the image, your triage action stamps it defect, and Jira Create issuefiles it with the drafting agent, preview confirmed, key logged. 2pm: “the export should also include the VAT column”. The triage action calls it a change request, so it goes to the backlog conversation instead of the defect list. 5pm: you select your closing notes, run your summary action (it sees the whole round through the session's compacted context), and publish the round status as a Confluence page. Nothing was re-typed twice, and the go/no-go meeting starts from a written record instead of memory.
Why not a tracker spreadsheet and a chatbot?
The classic setup is a UAT tracker spreadsheet plus a chatbot tab: copy the feedback out, paste it into the chat, re-explain the project, copy the result into the spreadsheet, then re-type it once more into Jira. Every hop loses context and adds minutes, multiplied by every item in the round. Khint collapses the hops: the classification runs on the text where it sits, the defect files to Jira from the same palette, and the sessioncarries the round's context so you never re-explain it. For the deep craft of one great bug report, see bug report writing for QA; for the one-off “error on screen → ticket” moment, see capturing an error into a ticket. This page is the round-level loop that sits above both. And if you are still writing the test cases, start from test cases from user stories.
Common questions
What does the triage action actually do?
It is a saved AI Action you author once on the Agents page: a prompt that reads a piece of raw tester feedback and returns it classified the way your project sorts things, for example defect, change request, or training gap, with a one-line rationale. You don't have to write the prompt from a blank page: type a description of what it should do and press Generate for a first draft, then Improve to refine it. From then on it runs on whatever text you have selected, in any app, behind Cmd+Shift+K.
How do tester screenshots become tickets?
Through Capture. When a tester sends a screenshot of an error dialog, run Extract text from the palette: you draw a region and the recognized text lands on your clipboard, ready to triage and file. If the screenshot needs interpreting rather than transcribing, Capture & ask opens a chat about the captured image, so you can ask what the error actually says before you write the defect. Both paths also log into your active Memory session.
Do I need Jira open to file the defects?
No. Once Atlassian is connected on the MCP page (email plus API token), the palette carries Jira rows you can run from any app: Create issue lets an agent draft the ticket from your triaged text (or use the text as-is), shows you a preview behind a confirm button, and returns the new issue key. Add comment and Change status cover the retest loop, so a fixed defect gets its comment and its status change without a tab switch.
Where does the feedback text go?
Your work stays on your machine. Memory sessions and their entries live in a local database and are not synced to the cloud, and an Action only ever sees the text you selected plus, if you opted in, the compacted context of your active session. For feedback that contains personal data, any Action can opt into PII redaction, which replaces emails and phone numbers with placeholders before the text is sent to the model.
Run your next UAT round from one shortcut
Free with 10 AI actions and 5 captures per day. No credit card. Capture the feedback, classify it, file the defects.