After a Genie space has been deployed, business users mark several responses as incorrect using the 'Fix it' feedback option. The author notices the patterns in the Monitoring tab. What must the author do to actually improve Genie's responses to those types of questions?
Answer & explanation
Correct answer: A
WHY A: Databricks documentation explicitly states that Genie's behavior does not change based on user feedback alone. Authors must use the feedback patterns they observe in the Monitoring tab to manually update instructions, example SQL queries, or trusted assets. The Monitoring tab identifies problem areas; actual improvement requires the author to apply changes. WHY NOT B: There is no 'Refresh and Retrain' option in the Configure menu; Genie does not fine-tune an LLM from user feedback within the space. Improvement requires author-driven instruction updates, not automated retraining. WHY NOT C: Marking responses as confirmed correct in the Monitoring tab does update the review status for business users, but it does not automatically teach Genie new SQL patterns; the author still must add or update instructions manually. WHY NOT D: Individual workspace-level user feedback is not submitted to Databricks as model improvement tickets; Genie space optimization is handled by the space author through configuration changes, not Databricks engineers. WHY NOT E: Deleting example queries and restarting the warehouse reduces the contextual guidance that helps Genie generate accurate SQL; this would likely worsen performance, not improve it. The correct action is to add or refine instructions for the problematic question categories.