Clerks regularly report chatbot misses during case work. How should the program use that feedback for continuous improvement?
Select an answer to reveal the explanation.
Short Explanation
Frontline misses are free product research. Drop them on a continuous-improvement backlog and chip away—don’t shrug, punish, or nuke the whole stack every time.
Full Explanation
When clerks regularly report chatbot misses during case work, the program should capture those misses in a continuous-improvement backlog and work them deliberately. Frontline reports are high-signal evidence about policy gaps, retrieval failures, tone issues, and edge cases that synthetic test sets miss. A managed backlog turns anecdotes into prioritized fixes with owners and acceptance criteria. Ignoring frontline feedback because models improve automatically fails because production civic quality does not self-heal without data pipelines, evaluation, and content/process updates. Punishing clerks who report misses to discourage negativity fails by destroying the sensing system that protects residents. Restarting the entire platform from scratch after every single miss fails by confusing continuous improvement with chaotic rebuilds and burning capacity needed for targeted remediation. Exam caveat: both "ignore" and "rip-and-replace" extremes appear as distractors; disciplined backlog management is the sustainment pattern. Operational check: create a simple miss intake form for clerks, triage into the AI improvement backlog weekly, link fixes to releases, and report backlog burn-down beside quality sample trends.