Patients downvote a hospital portal agent, and the comments say the prep sheet it quotes is outdated. A teammate wants to retune the model immediately. What should the architect do first?
Select an answer to reveal the explanation.
Short Explanation
When the comment cards say the handout is out of date, rewriting the teacher's notes will not fix the handout. You triage those comments and correct the prep sheet before you decide the model needs retuning. Jumping straight to a new model leaves the stale sheet in the answers.
Full Explanation
User feedback is evidence about which step failed, and it has to be read before anyone retunes a model. These comments name an outdated prep sheet, so the first move is to triage that feedback and correct the sheet the agent quotes. Retuning the model on the downvote text teaches it to dodge words patients used, while the stale sheet remains the source of the next answer. Deleting the stored feedback clears the downvote count and destroys the clue that pointed at the sheet. Replacing the portal agent with a custom vision model treats a document-freshness complaint as an image problem and still leaves the prep sheet wrong. After the current sheet is in place, a new sample of questions shows whether any model behavior is still off. That second look is how you decide on retuning, and it comes after the content fix. Exam caveat: a downvote is a pointer, not a training label, until someone has sorted what the comment actually complains about. Operational check: update the prep sheet, ask the same patient question again, and confirm the reply follows the current sheet before any tuning job is opened.