A disaster-relief non-profit is reviewing several proposed AI projects. Select the TWO proposals below where AI is NOT the appropriate solution, based on the descriptions given. (Select TWO.)

Autonomously deciding which individual disaster survivors receive limited emergency housing assistance, without human review, based on incomplete and inconsistent field-collected data, is inappropriate because it removes human oversight from a high-stakes, ethically sensitive resource-allocation decision built on unreliable data, exactly the kind of decision that needs human judgment. Drafting a one-time internal memo, a task performed only once and never repeated, where writing it manually would take a staff member fifteen minutes, is inappropriate because the cost of building or configuring an AI solution for a single, brief, never-repeated task exceeds any benefit gained; a manual approach is far more efficient. Summarizing lengthy incident reports, routing volunteer sign-up emails by stated location, and detecting anomalies in expense report submissions across hundreds of field offices are all reasonable AI use cases: routing based on stated criteria, summarizing high volumes of text, and detecting anomalies across large datasets are exactly the kinds of repeatable, pattern-based, moderate-stakes tasks where AI adds real value. The general principle: AI is inappropriate for high-stakes decisions lacking human oversight and reliable data, and for one-off, low-effort tasks where automation costs more than it saves, not for repeatable, pattern-based tasks at scale.