Your team, thoughtfully built.
How to Triage Website Feedback With AI
To triage website feedback with AI, keep what the visitor saw apart from what you think caused it. Ask for each report as four lines: what was observed, what could explain it, what is still missing, and the one check to run next. Then rank by what the visitor was stopped from doing.
The example is two made-up reports about a workshop booking page.
Keep the report as it arrived
- "The booking button did nothing on my phone." The screenshot shows the button. It does not show the browser, the page address or what happened after the tap.
- "The arrival address is hard to read."
Two reports, and nothing yet that says they share a cause. Give it the exact wording, the date, the page and what the visitor was trying to do. Where you don't know something, say you don't know.
The request
The request: "Triage the two reports below. For each, separate what the visitor observed, possible causes, what evidence is missing and the smallest next check. Rank them by what the visitor was blocked from doing. Keep the original wording. Do not change the live site, and do not tell anyone to clear their cache as a guess."
What comes back
- Booking button, first: the visitor could not book. Next check: tap the button on a phone and see what the page does.
- Address, second: open the page at phone width and compare it with the approved text.
- Not known: which browser, and whether there was a network error.
- Done when: a real booking goes through on a phone. Not when the button looks right.
The mistakes to watch for
A confident diagnosis from a screenshot. "This is a caching issue" sounds like expertise. With a cropped image and no browser details it is a guess, and acting on it wastes the visitor's time and yours.
One fix claimed twice. Suppose the check finds the address in white text on a white background. Fix that contrast and look at it in light and dark. That fix says nothing about the booking button, which is still not investigated.
Close the loop
For each fix, note what changed and how you checked it. If you could not test on the visitor's actual phone, say so. Keep the report and its resolution together, so the next person who sees the same complaint finds the answer.
Run it in your Staffmor office
Open the office on the computer running it with its Start Staffmor launcher. If you already have a paired web office, open it here. Paste this request and the source material into the conversation. Use your approved Codex connection, or ask your open Claude Code session to "Check Staffmor for work." A saved request is not a completed result. Review the answer or file when it comes back and reply with corrections. Ask for the triage first; investigate one report at a time after that.
Staffmor sorts the evidence and can trace an issue in the files it can read. Changing your live site stays with you. New to Staffmor? Set up your office with Codex or Claude Code and an empty folder. Use your own supported AI account and approve the download, office activation and local folder/network access during setup. Claude Code pickup is manual; keep that local session open.
Related: How to summarize product feedback with AI · How to set up an AI agent team