Exam Bundle

That observation is a valid indicator for “An optional JSON property is missing and the query must not crash” when using Use null-safe path extraction with explicit default handling. The scenario instead tests “Raw JSON strings should be parsed as structured variant values before querying fields”; the relevant result is Subsequent path queries return correctly typed or cast results.

This signal belongs to the adjacent case “A variable list of event objects must be expanded to one row per contained item,” whose remedy is Use supported variant explosion or array expansion functions. The observed result for the present requirement is Subsequent path queries return correctly typed or cast results.

This signal belongs to the adjacent case “An analyst needs to read a nested customer identifier from a VARIANT value,” whose remedy is Use supported variant-path extraction or accessor expressions. The observed result for the present requirement is Subsequent path queries return correctly typed or cast results.

Mask or exclude protected fields and use approved governed model access is appropriate when “A personal-data column must not be sent to an unapproved inference endpoint”; it would be verified by The request payload contains no prohibited sensitive attributes. It does not directly satisfy “A curated text column needs a supported model-backed SQL enrichment for a chosen task,” whose relevant acceptance evidence is The enrichment result returns the expected result schema for sampled records.

Evaluate outputs against curated labeled test records and track a quality metric belongs to the separate case “AI-generated labels are inconsistent and need acceptance evidence,” where Regression tests detect a change in label accuracy or format compliance would be expected. That differs from the condition in this stem: “A curated text column needs a supported model-backed SQL enrichment for a chosen task.”

The alternative Validate and parse the model response against an explicit output contract has a distinct purpose—A model response is free text but downstream processing requires a strict schema. It would not establish the required outcome (A curated text column needs a supported model-backed SQL enrichment for a chosen task); the necessary verification concerns The enrichment result returns the expected result schema for sampled records.

That observation is a valid indicator for “A personal-data column must not be sent to an unapproved inference endpoint” when using Mask or exclude protected fields and use approved governed model access. The scenario instead tests “A curated text column needs a supported model-backed SQL enrichment for a chosen task”; the relevant result is The enrichment result returns the expected result schema for sampled records.

Dashboard queries read stable stored enrichment columns would demonstrate the separate action Materialize validated model outputs in a governed table on a controlled schedule, which addresses “AI results should not be recomputed for every dashboard query.” It does not verify the requirement “A curated text column needs a supported model-backed SQL enrichment for a chosen task” following Use an authorized `ai_query` call or supported AI SQL function with defined inputs.

This signal belongs to the adjacent case “A model response is free text but downstream processing requires a strict schema,” whose remedy is Validate and parse the model response against an explicit output contract. The observed result for the present requirement is The enrichment result returns the expected result schema for sampled records.

The alternative Evaluate outputs against curated labeled test records and track a quality metric has a distinct purpose—AI-generated labels are inconsistent and need acceptance evidence. It would not establish the required outcome (AI results should not be recomputed for every dashboard query); the necessary verification concerns Dashboard queries read stable stored enrichment columns.

Use an authorized `ai_query` call or supported AI SQL function with defined inputs belongs to the separate case “A curated text column needs a supported model-backed SQL enrichment for a chosen task,” where The enrichment result returns the expected result schema for sampled records would be expected. That differs from the condition in this stem: “AI results should not be recomputed for every dashboard query.”

This action solves a different issue: “A model response is free text but downstream processing requires a strict schema.” Its normal signal is Malformed model output is quarantined rather than entering the curated table. Here the requested behavior is “AI results should not be recomputed for every dashboard query,” so that signal is not sufficient.

That observation is a valid indicator for “A curated text column needs a supported model-backed SQL enrichment for a chosen task” when using Use an authorized `ai_query` call or supported AI SQL function with defined inputs. The scenario instead tests “AI results should not be recomputed for every dashboard query”; the relevant result is Dashboard queries read stable stored enrichment columns.

This signal belongs to the adjacent case “AI-generated labels are inconsistent and need acceptance evidence,” whose remedy is Evaluate outputs against curated labeled test records and track a quality metric. The observed result for the present requirement is Dashboard queries read stable stored enrichment columns.

Malformed model output is quarantined rather than entering the curated table pertains to a different control (Validate and parse the model response against an explicit output contract) and a different need (A model response is free text but downstream processing requires a strict schema). It cannot establish that Materialize validated model outputs in a governed table on a controlled schedule met the stated goal.

The alternative Use an authorized `ai_query` call or supported AI SQL function with defined inputs has a distinct purpose—A curated text column needs a supported model-backed SQL enrichment for a chosen task. It would not establish the required outcome (A model response is free text but downstream processing requires a strict schema); the necessary verification concerns Malformed model output is quarantined rather than entering the curated table.

Mask or exclude protected fields and use approved governed model access belongs to the separate case “A personal-data column must not be sent to an unapproved inference endpoint,” where The request payload contains no prohibited sensitive attributes would be expected. That differs from the condition in this stem: “A model response is free text but downstream processing requires a strict schema.”

Evaluate outputs against curated labeled test records and track a quality metric belongs to the separate case “AI-generated labels are inconsistent and need acceptance evidence,” where Regression tests detect a change in label accuracy or format compliance would be expected. That differs from the condition in this stem: “A model response is free text but downstream processing requires a strict schema.”

That observation is a valid indicator for “A curated text column needs a supported model-backed SQL enrichment for a chosen task” when using Use an authorized `ai_query` call or supported AI SQL function with defined inputs. The scenario instead tests “A model response is free text but downstream processing requires a strict schema”; the relevant result is Malformed model output is quarantined rather than entering the curated table.

This signal belongs to the adjacent case “A personal-data column must not be sent to an unapproved inference endpoint,” whose remedy is Mask or exclude protected fields and use approved governed model access. The observed result for the present requirement is Malformed model output is quarantined rather than entering the curated table.

This signal belongs to the adjacent case “AI-generated labels are inconsistent and need acceptance evidence,” whose remedy is Evaluate outputs against curated labeled test records and track a quality metric. The observed result for the present requirement is Malformed model output is quarantined rather than entering the curated table.

Validate and parse the model response against an explicit output contract belongs to the separate case “A model response is free text but downstream processing requires a strict schema,” where Malformed model output is quarantined rather than entering the curated table would be expected. That differs from the condition in this stem: “A personal-data column must not be sent to an unapproved inference endpoint.”

The alternative Use an authorized `ai_query` call or supported AI SQL function with defined inputs has a distinct purpose—A curated text column needs a supported model-backed SQL enrichment for a chosen task. It would not establish the required outcome (A personal-data column must not be sent to an unapproved inference endpoint); the necessary verification concerns The request payload contains no prohibited sensitive attributes.

This action solves a different issue: “AI-generated labels are inconsistent and need acceptance evidence.” Its normal signal is Regression tests detect a change in label accuracy or format compliance. Here the requested behavior is “A personal-data column must not be sent to an unapproved inference endpoint,” so that signal is not sufficient.

That observation is a valid indicator for “A curated text column needs a supported model-backed SQL enrichment for a chosen task” when using Use an authorized `ai_query` call or supported AI SQL function with defined inputs. The scenario instead tests “A personal-data column must not be sent to an unapproved inference endpoint”; the relevant result is The request payload contains no prohibited sensitive attributes.

Regression tests detect a change in label accuracy or format compliance pertains to a different control (Evaluate outputs against curated labeled test records and track a quality metric) and a different need (AI-generated labels are inconsistent and need acceptance evidence). It cannot establish that Mask or exclude protected fields and use approved governed model access met the stated goal.

That observation is a valid indicator for “A model response is free text but downstream processing requires a strict schema” when using Validate and parse the model response against an explicit output contract. The scenario instead tests “A personal-data column must not be sent to an unapproved inference endpoint”; the relevant result is The request payload contains no prohibited sensitive attributes.

This action solves a different issue: “AI results should not be recomputed for every dashboard query.” Its normal signal is Dashboard queries read stable stored enrichment columns. Here the requested behavior is “AI-generated labels are inconsistent and need acceptance evidence,” so that signal is not sufficient.

This action solves a different issue: “A curated text column needs a supported model-backed SQL enrichment for a chosen task.” Its normal signal is The enrichment result returns the expected result schema for sampled records. Here the requested behavior is “AI-generated labels are inconsistent and need acceptance evidence,” so that signal is not sufficient.

Mask or exclude protected fields and use approved governed model access is appropriate when “A personal-data column must not be sent to an unapproved inference endpoint”; it would be verified by The request payload contains no prohibited sensitive attributes. It does not directly satisfy “AI-generated labels are inconsistent and need acceptance evidence,” whose relevant acceptance evidence is Regression tests detect a change in label accuracy or format compliance.

This signal belongs to the adjacent case “A personal-data column must not be sent to an unapproved inference endpoint,” whose remedy is Mask or exclude protected fields and use approved governed model access. The observed result for the present requirement is Regression tests detect a change in label accuracy or format compliance.

This signal belongs to the adjacent case “A model response is free text but downstream processing requires a strict schema,” whose remedy is Validate and parse the model response against an explicit output contract. The observed result for the present requirement is Regression tests detect a change in label accuracy or format compliance.

That observation is a valid indicator for “AI results should not be recomputed for every dashboard query” when using Materialize validated model outputs in a governed table on a controlled schedule. The scenario instead tests “AI-generated labels are inconsistent and need acceptance evidence”; the relevant result is Regression tests detect a change in label accuracy or format compliance.