Question 60 A finance analyst asks for the current receivables total, while a planned RAG pipeline accepts only screenshots and has no authoritative ledger read. Why does the current approach fail under these conditions? Eligibility rules depend on jurisdiction and permit class; the model cannot ground a decision if required policy selectors are missing. A prose answer is not interchangeable with the numeric contract that dispatch software uses to prioritize exceptions. Field extraction turns varied documents into typed records; QA and chat-routing features solve different tasks. Ledger totals are live transactional facts; a screenshot-centric language model cannot guarantee present balances or precise aggregation. ExplanationCorrect Answer D:Define an authorized ledger-query input keyed by reporting date and entity, producing an amount, currency, freshness timestamp and source identifier. Other Answers A:Ledger totals are live transactional facts; a screenshot-centric language model cannot guarantee present balances or precise aggregation. B:Ledger totals are live transactional facts; a screenshot-centric language model cannot guarantee present balances or precise aggregation. C:That mechanism describes another problem: An accounts-payable team needs to pull invoice number, dates and totals from semistructured vendor documents, not answer conversational questions about them. It does not explain this case, where ledger totals are live transactional facts; a screenshot-centric language model cannot guarantee present balances or precise aggregation.
Question 61 A finance analyst asks for the current receivables total, while a planned RAG pipeline accepts only screenshots and has no authoritative ledger read. Which observation would establish that the requirement is now met? Replay cases spanning two districts and a missing permit type, checking the rule citations and explicit abstention for incomplete inputs. Compare parsed delay and confidence fields against held-out deliveries, with failures recorded for missing shipment identifiers. Compare extracted totals, dates and supplier IDs against labeled invoices and reject unsupported field guesses. Reconcile an answer for two closing dates against the SQL ledger and verify both the currency and source timestamp. ExplanationCorrect Answer D:Ledger totals are live transactional facts; a screenshot-centric language model cannot guarantee present balances or precise aggregation. Other Answers A:Reconcile an answer for two closing dates against the SQL ledger and verify both the currency and source timestamp. B:Reconcile an answer for two closing dates against the SQL ledger and verify both the currency and source timestamp. C:Reconcile an answer for two closing dates against the SQL ledger and verify both the currency and source timestamp.
Question 62 A finance analyst asks for the current receivables total, while a planned RAG pipeline accepts only screenshots and has no authoritative ledger read. What lasting control best prevents recurrence of the identified issue? Keep a versioned extraction contract, labeled test set and exception-review queue for new document templates. Version the serving schema and regression-test numeric field types whenever the estimation model changes. Require a business-approved input/output contract and missing-field acceptance cases before the first model implementation. Document the live system of record, row-level access rules and reconciliation tests in the use-case contract. ExplanationCorrect Answer D:Ledger totals are live transactional facts; a screenshot-centric language model cannot guarantee present balances or precise aggregation. Other Answers A:Ledger totals are live transactional facts; a screenshot-centric language model cannot guarantee present balances or precise aggregation. B:This control addresses a different exposure: A logistics dashboard wants delivery-delay predictions, but the proposed assistant returns a narrative instead of the expected delay_minutes and confidence fields. It would not prevent this failure, because ledger totals are live transactional facts; a screenshot-centric language model cannot guarantee present balances or precise aggregation. C:That ledger totals are live transactional facts; a screenshot-centric language model cannot guarantee present balances or precise aggregation.
Question 63 A multilingual benefits assistant is expected to explain exclusions, but the business goal omits the benefit plan year and desired output language. What is the most targeted corrective step for this observed failure? Use a Genie data-question workflow tied to governed semantic definitions and permissioned data, rather than selecting document-only Knowledge Assistant. Define typed permit_type, district and submitted_facts inputs, with an eligibility decision, evidence citations and an explicit insufficient_information outcome. Translate the business question into a typed shipment identifier, event-history inputs and numeric delay_minutes plus calibrated confidence outputs. Require plan_id, effective_date and locale as validated inputs and provide a cited explanation with exclusions and unknown-state flags. ExplanationCorrect Answer D:The action targets the stated cause: Coverage and document validity vary with plan version; language requirements change presentation, not the authoritative policy facts. Check the repair by: test an expired plan, two locales and an absent effective date; verify citations resolve to the correct plan version. Other Answers A:Coverage and document validity vary with plan version; language requirements change presentation, not the authoritative policy facts. B:Coverage and document validity vary with plan version; language requirements change presentation, not the authoritative policy facts. C:Coverage and document validity vary with plan version; language requirements change presentation, not the authoritative policy facts.
Question 64 A multilingual benefits assistant is expected to explain exclusions, but the business goal omits the benefit plan year and desired output language. What is the decisive mechanism behind the observed result? Eligibility rules depend on jurisdiction and permit class; the model cannot ground a decision if required policy selectors are missing. Coverage and document validity vary with plan version; language requirements change presentation, not the authoritative policy facts. Structured questions require query planning over trusted table semantics and governed data access rather than arbitrary similarity retrieval. A prose answer is not interchangeable with the numeric contract that dispatch software uses to prioritize exceptions. ExplanationCorrect Answer B:Require plan_id, effective_date and locale as validated inputs and provide a cited explanation with exclusions and unknown-state flags. Other Answers A:Coverage and document validity vary with plan version; language requirements change presentation, not the authoritative policy facts. C:That mechanism describes another problem: A department wants natural-language analysis of governed business tables, not open-ended document search, and asks which Databricks product family fits the need. It does not explain this case, where coverage and document validity vary with plan version; language requirements change presentation, not the authoritative policy facts. D:Coverage and document validity vary with plan version; language requirements change presentation, not the authoritative policy facts.
Question 65 A multilingual benefits assistant is expected to explain exclusions, but the business goal omits the benefit plan year and desired output language. Which verification is strongest for this particular failure mode? Test an expired plan, two locales and an absent effective date; verify citations resolve to the correct plan version. Replay cases spanning two districts and a missing permit type, checking the rule citations and explicit abstention for incomplete inputs. Compare parsed delay and confidence fields against held-out deliveries, with failures recorded for missing shipment identifiers. Ask a test business question and verify the executed query uses approved tables, metrics and caller privileges. ExplanationCorrect Answer A:Coverage and document validity vary with plan version; language requirements change presentation, not the authoritative policy facts. Other Answers B:Test an expired plan, two locales and an absent effective date; verify citations resolve to the correct plan version. C:Test an expired plan, two locales and an absent effective date; verify citations resolve to the correct plan version. D:Test an expired plan, two locales and an absent effective date; verify citations resolve to the correct plan version.
Question 66 A multilingual benefits assistant is expected to explain exclusions, but the business goal omits the benefit plan year and desired output language. What change to engineering governance would reduce this risk over time? Require a business-approved input/output contract and missing-field acceptance cases before the first model implementation. Maintain a signed use-case specification listing all identity, temporal, language and citation constraints. Version the serving schema and regression-test numeric field types whenever the estimation model changes. Curate data semantics and test row-level security before enabling business-facing natural-language questions. ExplanationCorrect Answer B:Coverage and document validity vary with plan version; language requirements change presentation, not the authoritative policy facts. Other Answers A:This control addresses a different exposure: A city help desk requests permit eligibility answers, but the intake contract contains only a free-text complaint and no permit type or governing district. It would not prevent this failure, because coverage and document validity vary with plan version; language requirements change presentation, not the authoritative policy facts. C:That coverage and document validity vary with plan version; language requirements change presentation, not the authoritative policy facts. D:Coverage and document validity vary with plan version; language requirements change presentation, not the authoritative policy facts.
Question 67 A purchasing agent searches policy, calculates a vendor discount and issues a purchase order before checking the user's spending authority. Which engineering action most directly addresses this specific requirement? Treat the API response as untrusted data, cap iterations and let the planner choose only authorized tools from a fixed policy. Redact internal identifiers before external research and route private diagnostics only through approved authenticated internal tools. Order the workflow as authenticate, retrieve authorized terms, calculate exact totals, obtain purchase approval and only then call the write tool. Define an authorized ledger-query input keyed by reporting date and entity, producing an amount, currency, freshness timestamp and source identifier. ExplanationCorrect Answer C:The action targets the stated cause: Tool calls are consequential side effects; neither a model prediction nor a retrieved policy paragraph can substitute for an authorization boundary. Check the repair by: replay an over-limit order and verify no purchase API call occurs before a recorded approval event. Other Answers A:Tool calls are consequential side effects; neither a model prediction nor a retrieved policy paragraph can substitute for an authorization boundary. B:Tool calls are consequential side effects; neither a model prediction nor a retrieved policy paragraph can substitute for an authorization boundary. D:Tool calls are consequential side effects; neither a model prediction nor a retrieved policy paragraph can substitute for an authorization boundary.
Question 68 A purchasing agent searches policy, calculates a vendor discount and issues a purchase order before checking the user's spending authority. What architectural constraint is most directly responsible? Tool calls are consequential side effects; neither a model prediction nor a retrieved policy paragraph can substitute for an authorization boundary. Ledger totals are live transactional facts; a screenshot-centric language model cannot guarantee present balances or precise aggregation. A tool result cannot confer authority on itself; missing loop limits can turn injected instructions into runaway tool execution. External search receives the query as data; a chained prompt does not undo sensitive information already transmitted. ExplanationCorrect Answer A:Order the workflow as authenticate, retrieve authorized terms, calculate exact totals, obtain purchase approval and only then call the write tool. Other Answers B:That mechanism describes another problem: A finance analyst asks for the current receivables total, while a planned RAG pipeline accepts only screenshots and has no authoritative ledger read. It does not explain this case, where tool calls are consequential side effects; neither a model prediction nor a retrieved policy paragraph can substitute for an authorization boundary. C:Tool calls are consequential side effects; neither a model prediction nor a retrieved policy paragraph can substitute for an authorization boundary. D:Tool calls are consequential side effects; neither a model prediction nor a retrieved policy paragraph can substitute for an authorization boundary.
Question 69 A purchasing agent searches policy, calculates a vendor discount and issues a purchase order before checking the user's spending authority. What validation evidence would most specifically confirm the resolution? Feed a malicious refresh directive through a mocked tool and verify one bounded attempt, safe termination and trace evidence. Reconcile an answer for two closing dates against the SQL ledger and verify both the currency and source timestamp. Replay an over-limit order and verify no purchase API call occurs before a recorded approval event. Inspect outbound tool payloads from synthetic private tickets and confirm no serial, address or internal notes leave the trusted boundary. ExplanationCorrect Answer C:Tool calls are consequential side effects; neither a model prediction nor a retrieved policy paragraph can substitute for an authorization boundary. Other Answers A:Replay an over-limit order and verify no purchase API call occurs before a recorded approval event. B:Replay an over-limit order and verify no purchase API call occurs before a recorded approval event. D:Replay an over-limit order and verify no purchase API call occurs before a recorded approval event.
Question 70 A purchasing agent searches policy, calculates a vendor discount and issues a purchase order before checking the user's spending authority. What prevention measure is most targeted to this underlying cause? Apply per-tool data classifications, payload allowlists and outbound-call logging to agent orchestration. Document the live system of record, row-level access rules and reconciliation tests in the use-case contract. Make spending thresholds and approval gates server-side controls with an immutable tool-call audit trail. Use maximum tool-call budgets and policy checks independent of the language model's generated plans. ExplanationCorrect Answer C:Tool calls are consequential side effects; neither a model prediction nor a retrieved policy paragraph can substitute for an authorization boundary. Other Answers A:That tool calls are consequential side effects; neither a model prediction nor a retrieved policy paragraph can substitute for an authorization boundary. B:Tool calls are consequential side effects; neither a model prediction nor a retrieved policy paragraph can substitute for an authorization boundary. D:This control addresses a different exposure: A support agent repeatedly calls a pricing API because a tool response contains an instruction telling it to refresh itself indefinitely. It would not prevent this failure, because tool calls are consequential side effects; neither a model prediction nor a retrieved policy paragraph can substitute for an authorization boundary.
